CloudCom 2010 Papers and Posters/Demos

Auditing, Monitoring and Scheduling

Initial Findings for Provisioning Variation in Cloud Computing

[slides]
Mohammed Suhail Rehman and Majd F. Sakr
Carnegie Mellon University in Qatar

Abstract: Cloud computing offers a paradigm shift in management of computing resources for large-scale applications. Using the Infrastructure-as-a-service (IaaS) cloud computing model, users today can request dynamically provisioned, virtualized resources such as CPU, memory, disk, and network access in the form of virtualized resources. The client typically requests resources based on computational needs and pays for resource instances based on their capacity and time utilized. Mapping these virtual resource requests to physical hardware could vary for identical requests. This can potentially cause variations in the performance of applications deployed on such resources.

The performance of the application can vary according to the physical layout of the provisioned hardware (the number of virtual machines (VMs), the size/configuration of the VMs and the inter-VM locality). In this paper, we study the effects of this ”provisioning variation” and its impact on application performance using suitable benchmarks as well demonstrate their effect on a few MapReduce workloads. Our initial findings indicate that provisioning variation can impact performance by a factor of 5 primarily due to I/O contention.

A Hybrid and Secure Mechanism to Execute Parameter Survey Applications on Local and Public Cloud Resources

[slides]
Hao Sun and Kento Aida
Tokyo Institute of Technology

Abstract: Abstract: A parameter survey application (PSA) is a typical application running on high-performance computing (HPC) systems. A PSA consists of a lot of independent tasks with different input parameters that are executed in parallel on different CPU cores. Infrastructure-as-a-Service Cloud (IaaS Cloud) is expected to be used as an HPC infrastructure to run PSAs, and some reports have discussed hybrid execution mechanisms to utilize both local resources and Iaas Cloud. However, users still have security problems in running applications with confidential data on IaaS Cloud. We propose a hybrid and secure execution mechanism to run PSAs utilizing both local computing resources with a batch scheduler and IaaS Cloud. The proposed mechanism utilizes both local resources and Iaas Cloud to meet the deadline of user applications. We conducted experiments running a natural language processing application, which uses machine learning to detect abusive language on Internet bulletin board systems. The experimental results showed that the proposed mechanism effectively allocated resources and met the deadlines of the user application.

CloudView: Describe and Maintain Resource View in Cloud

[slides]
Dehui Zhou, Liang Zhong, Tianyu Wo and Junbin Kang
School of Computer Science and Engineering, Beihang University

Abstract: Abstract: Resource view is the user defined table to provide specific view on resource status in cloud computing environment. It provides a convenient way to retrieve resource data for applications at infrastructure, platform and service layers. But the description and maintenance of these diverse resource views are inconvenient and dramatically difficult due to massive, heterogeneous and dynamic characteristics of the cloud resources involved. In this paper we present a resource view description scheme RQL and the corresponding system CloudView to address these difficulties. RQL provides users a scheme to specify the data processing flow from resource raw data collected to resource view data objected. By constructing data processing acyclic graph based on view definitions and using basic routines, view maintenance mechanism update user defined resource views automatically and periodically. CloudView use a centralized scheduler to distribute maintenance jobs to a set of scalable worker nodes. It leverages distributed key-value database to store view data. Compared to related resource monitoring and discovering systems, CloudView is flexible in application oriented view description. Experiments show it updates typical user defined views with desired performance.

VDBench: A Benchmarking Toolkit for Thin-client based Virtual Desktop Environments

[slides]
Alex Berryman, Prasad Calyam, Matthew Honigford and Albert Lai
The Ohio State University, VMWare

Abstract: The recent advances in thin client devices and the push to transition users' desktop delivery to cloud environments will eventually transform how desktop computers are used today. The ability to measure and adapt the performance of virtual desktop environments is a major challenge for "virtual desktop cloud" service providers. In this paper, we present the "VDBench" toolkit that uses a novel methodology and related metrics to benchmark thin-client based virtual desktop environments in terms of scalability and reliability. We also describe how we used a VDBench instance to benchmark the performance of: (a) popular user applications (Spreadsheet Calculator, Internet Browser, Media Player, Interactive Visualization), (b) TCP/UDP based thin client protocols (RDP, RGS, PCoIP), and (c) remote user experience (interactivity, perceived video quality), under a variety of system load and network health conditions. Our results can help service providers to mitigate over-provisioning in sizing virtual desktop resources, and guesswork in thin client protocol configurations, and thus obtain significant cost savings while simultaneously fostering satisfied customers.

Using Global Behavior Modeling to Improve QoS in Cloud Data Storage Services

[slides]
Jesus Montes, Bogdan Nicolae, Gabriel Antoniu, Alberto Sanchez and Maria S. Perez
University of Rennes 1/IRISA, INRIA Rennes, Universidad Rey Juan Carlos, Universidad Politecnica de Madrid

Abstract: The cloud computing model aims to make large-scale data-intensive computing affordable even for users with limited financial resources, that cannot invest into expensive infrastructures necessary to run them. In this context, MapReduce is emerging as a highly scalable programming paradigm that enables high-throughput data-intensive processing as a cloud service. Its performance is highly dependent on the underlying storage service, responsible to efficiently support massively parallel data accesses by guaranteeing a high throughput under heavy access concurrency. In this context, quality of service plays a crucial role: the storage service needs to sustain a stable throughput for each individual access, in addition to achieving a high aggregated throughput under concurrency.

In this paper we propose a technique to address this problem using component monitoring, application-side feedback and behavior pattern analysis to automatically infer useful knowledge about the causes of poor quality of service and provide an easy way to reasoning about potential improvements. We apply our proposal to BlobSeer, a representative data storage service specifically designed to achieve high aggregated throughput and show through extensive experimentation substantial improvements in the stability of individual data read accesses under MapReduce workloads.

Bag-of-Tasks Scheduling under Budget Constraints

[slides]
Ana-Maria Oprescu and Thilo Kielmann
Vrije Universiteit

Abstract: The new ways of doing science, rooted on the unprecedented processing, communication and storage infrastructures that became available to scientists, are collectively called e-Science. Many research labs now need non-trivial computational power during some periods of time, when they need to run e-science applications. Grid and voluntary computing are well-established solutions that cater to this need, but are not accessible for all labs and institutions. Besides, there is an uncertainty about the future amount of resources that will be available in such infrastructures, which prevents the researchers from planning their activities to guarantee that deadlines will be met. With the emergence of the cloud computing paradigm come new opportunities. One possibility is to run e-science activities at resources acquired on-demand from cloud providers.

However, although very low, there is a cost associated with the usage of cloud resources. Besides that, the amount of resources that can be simultaneously acquired is, in practise, limited. Another possibility is the not new idea of composing hybrid infrastructures in which the huge amount of computational resources shared by the grid infrastructures are used whenever possible and extra capacity is acquired from cloud computing providers. We here investigate how to schedule e-science activities in such hybrid infrastructures so that deadlines are met and costs are reduced. Our preliminary results indicate that online strategies can be used as an alternative to achieve better results than a strategy which prematurely acquires resources to run the application.

Dynamic request allocation and scheduling for context aware applications subject to a percentile response time SLA in a distributed cloud

[slides]
Keerthana Boloor, Rada Chirkova, Yannis Viniotis and Tiia Salo
North Carolina State University, IBM

Abstract: We consider geographically distributed datacenters forming a collectively managed cloud computing system, hosting multiple Service Oriented Architecture (SOA) based context aware applications, each subject to Service Level Agreements (SLA). The Service Level Agreements for each context aware application require the response time of a certain percentile of the input requests to be less than a specified value for a profit to be charged by the cloud provider. We present a novel approach of data-oriented dynamic service-request allocation with gi-FIFO scheduling, in each of the geographically distributed datacenters, to globally maximize the profit charged by the cloud computing system. Our evaluation shows that our dynamic scheme far outperforms the commonly deployed static allocation with either First in First Out (FIFO) or Weighted Round Robin (WRR) scheduling.

Usage Patterns to Provision for Time Critical Scientific Experimentation in Clouds

[slides]
Eran Chinthaka Withana and Beth Plale
Indiana University

Abstract: Driven by the need to provision resources on demand, scientists have turned to commercial and research test-bed Cloud computing resources to run their scientific experiments. Job scheduling on cloud computing resources, unlike earlier platforms, is a balance between throughput and cost of executions. Within this context, we posit that usage patterns can improve the job execution, because these patterns allow a system to plan, stage and make better scheduling decisions. This paper introduces a novel approach to utilization of user patterns drawn from knowledge-based techniques, to improve execution across a series of active workflows and jobs in cloud computing environments. Using empirical analysis we establish the accuracy of our prediction approach for two different workloads and demonstrate how this knowledge can be used to improve job executions.

A Novel Parallel Traffic Control Mechanism for Cloud Computing

[slides]
Zheng Li and Nenghai Yu
University of Science and Technology of China

Abstract: In this paper a novel parallel network traffic control mechanism for cloud computing is proposed based on the packet scheduler HTB (Hierarchical Token Buckets). The idea of bandwidth borrowing in HTB makes it suitable for large scale high-performance scenario such as cloud computing. However, the capacity of current HTB is only 0.5Gbps, which could not afford the high traffic rate and user concurrency in the cloud. In this paper, a parallel HTB mechanism is proposed: new algorithms are designed to reduce concurrency in HTB key structure access. Then, the usage of lock-free FIFOs parallelizes HTB into a 2-stage pipeline on the multi-core architecture. This parallel HTB could not only increase the processing rate, but also keep a well performance on stability, which is important for cloud computing. The simulation results prove that parallel HTB’s capacity can easily break 1Gbps, and reach 2Gbps at most, increasing the performance as high as 300% compared to the traditional HTB. This makes it an effective network traffic control mechanism for cloud computing. Moreover, our work is the first one to make HTB run in the pipelined fashion on a multi-core processor.

Fine-grained Data Access Control Systems with User Accountability in Cloud Computing

[slides]
Jin Li, Gansen Zhao and Chunming Rong
Guangzhou University China, South China Normal University, Stavanger University Norway

Abstract: Cloud computing is an emerging computing paradigm in which IT resources and capacities are provided as
services over the Internet. Promising as it is, this paradigm also brings forth new challenges for data security
and access control when users outsource sensitive data for sharing on cloud servers, which are likely outside of the
same trust domain of data owners. To maintain the confidentiality of sensitive user data against untrusted servers, existing work
usually apply cryptographic methods by disclosing data decryption keys only to authorized users. However, in doing so,
these solutions inevitably introduce heavy computation overhead on the data owner for key distribution and data
management when fine-grained data access control is desired, and thus do not scale well.

In this paper, we present a way to implement scalable and fine-grained access control systems based on attribute-based encryption (ABE). For the purpose of secure access control in cloud computing, the prevention of illegal key sharing among colluding users is missing from the existing access control systems based on ABE. This paper addresses this challenging open issue by defining and enforcing access policies based on data attributes and implementing user accountability by using traitor tracing. Furthermore, both the user grant and revocation are efficiently supported by using the broadcast encryption technique.Extensive analysis shows that the proposed scheme is highly efficient and provably secure under existing security models.

Computer Architectures

LEEN: Locality/Fairness- aware key partitioning for MapReduce in the Cloud

[slides]
Shadi Ibrahim, Hai Jin, Lu Lu, Bingsheng He, Li Qi and Song Wu
Huazhong University of Science and Technology, Nanyang Technological University, China Development Bank

Abstract: This paper investigates the problem of Partitioning Skew in MapReduce-like system. Our studies with Hadoop, widely used MapReduce implementation, demonstrate, in the presence of partitioning skew, a huge amount of data transfer during the shuffle phase incurring network congestion and leading to significant unfairness on the reduce input among different datanodes. As a result of partitioning skew, the applications running on MapReduce experience performance degradation due to the computation skew, particularly in reduce phase.

We developed a novel algorithm named LEEN for locality-aware and fairness-aware key partitioning in MapReduce. LEEN embraces an asynchronous Map and Reduce scheme. All buffered intermediate keys are partitioned according to their frequencies and the fairness of the expected data distribution after the shuffle phase. We have integrated LEEN into Hadoop-0.18.0. Our experiments demonstrate that LEEN can efficiently achieve higher locality in MapReduce-like systems and reduce the amount of shuffled data and more importantly, guarantee fair distribution of the reduce inputs. As a result, LEEN achieves a performance improvement of up to 40% on different metrics that may reflect different applications in the Cloud.

Towards a Reference Architecture for Semantically Interoperable Clouds

[slides]
Nikolaos Loutas, Vassilios Peristeras, Thanassis Bouras, Eleni Kamateri, Dimitrios Zeginis and Konstantinos Tarabanis
Center for Research and Technology Hellas (CERTH), DERI Galway, UBITECH, University of Macedonia

Abstract: This paper focuses on the emerging problem of semantic interoperability between heterogeneous cooperating Cloud platforms. We try to pave the way towards a Reference Architecture for Semantically Interoperable Clouds (RASIC). To this end, three fundamental and complementary computing paradigms, namely Cloud computing, Service Oriented Architectures and lightweight semantics are used and provide the main building blocks. The open, generic Reference Architecture for Semantically Interoperable Clouds introduces a scalable, reusable and transferable approach for facilitating the design, deployment and execution of resource intensive SOA services on top of semantically interlinked Clouds. In order to support the development of semantically interoperable Cloud systems based on RASIC, the model of a common Cloud API is also specified.

Performing Large Science Experiments within a Cloud Architecture: Pitfalls and Solutions

[slides]
Wei Lu, Jared Jackson, Jaliya Ekanayake, Roger Barga and Nelson Araujo
Microsoft

Abstract: Carrying out science at extreme scale is the next generational challenge facing the broad field of scientific research. Recent advances in general access to large data and compute centers through cloud computing provides the potential for an increasing number of researchers to have the ability to tackle new challenges in their field. Unfortunately barriers of complexity remain for researchers untrained in cloud programming. In this paper we examine how cloud based architectures can be made to solve largescale research experiments in a manner that is easily accessible for researchers with limited programming experience attempting to utilize their existing computational tools. We examine the top challenges identified in our own large-scale science experiments running on Windows Azure platform and then describe a Cloud-based parameter sweep prototype (dubbed Cirrus) which provides a framework of solutions for each challenge.

Exploratory Project: State of the Cloud, from University of Michigan and Beyond

[slides]
Traci Ruthkoski
University of Michigan

Abstract: As data center costs rise and space availability diminishes many organizations are investigating the viability of cloud computing for research use. Yet the majority of research investigators have not readily embraced cloud capability, regardless of the potential cost savings. Through interviews, case studies, and up-to-the-minute blog posts from top experts, it is possible to extract a basic framework of barriers that dampen widespread cloud adoption. Insight gained through examining these barriers can then be used to design an organizational strategic plan to build a cloud-enhanced campus cyberinfrastructure.

Abstractions for Loosely-Coupled and Ensemble-based Simulations on Azure

[slides]
Shantenu Jha and Andre Luckow
CCT / LSU

Abstract: Azure is an emerging cloud platform developed and operated by Microsoft. It provides a range of abstractions and building blocks for creating scalable and reliable scientific applications. In this paper we investigate the applicability of the Azure abstractions to the well-known class of loosely coupled and ensemble-based applications. We propose the BigJob API as a novel abstraction for managing groups of Azure worker roles and for remotely executing tasks on them. We demonstrate that Azure enhanced with BigJob functionality provides performance comparable to other grid and cloud offerings loosely-coupled applications.

Sustainable Network Resource Management System for Virtual Private Clouds

[slides]
Takahiro Miyamoto, Michiaki Hayashi and Kosuke Nishimura
KDDI R&D Laboratories Inc.

Abstract: To satisfy the requirement of secure isolation of Infrastructure–as-a-Service (IaaS) for enterprise customers, the virtual private clouds, which are separated from the others by using virtualization technologies, are deployed. However, the isolation with virtualization technologies can not avoid the affect of performance degradation, such as traffic congestion. Therefore, bandwidth-guaranteed virtual private clouds are needed for excluding unintentional and unwanted influence among multiple customers. In this paper, we propose a sustainable network resource management system (NRM) introducing a CHAnging MEchanism of software moduLE based on the cONtext (CHAMELEON) and a virtual network point for multipoint network provisioning. With the proposed mechanisms, we successfully demonstrated the sustainability of the NRM, which controls six kinds of network equipment without any modification itself.

Applying Twister to Scientific Applications

[slides]
Bingjing Zhang, Yang Ruan, Tak-Lon Wu, Judy Qiu, Adam Hughes and Geoffrey Fox
Indiana University

Abstract: Many scientific applications suffer from the lack of a unified approach to support the management in execution and inefficiency in processing large-scale data. Twister MapReduce Framework, which not only supports traditional MapReduce programming model but also extends it with iterations, tries to address these problems. This paper describes how Twister is applied to several kinds of scientific applications such as BLAST, MDS Interpolation and GTM Interpolation in non-iterative style and MDS without interpolation in iterative style. The results show the applicability of Twister to data parallel and EM algorithms with small overhead and increased efficiency.

Middleware level solutions for SPMD Applications in Grids and Clouds

[slides]
Brian Amedro, Francoise Baude, Fabrice Huet and Elton Mathias
INRIA, CNRS, I3S, UNSA

Abstract: Through the recent emergence of joint resource and network virtualization, dynamic composition and provisioning of time-limited and isolated virtual infrastructures is now possible. One other benefit of infrastructure virtualization is the capability of transparent reliability provisioning (reliability becomes a service provided by the infrastructure). In this context, we discuss the motivations and gains of introducing customizable reliability of virtual infrastructures when executing large-scale distributed applications, and present a framework to specify, allocate and deploy virtualized infrastructure with reliability capabilities. An approach to efficiently specify and control the reliability at runtime is proposed. We illustrate these ideas by analyzing the introduction of reliability at the virtual-infrastructure level on a real application. Experimental results, obtained with an actual medical-imaging application running in virtual infrastructures provisioned in the experimental large-scale Grid'5000 platform, show the benefits of the virtualization of reliability.

Attaching Cloud Storage to a Campus Grid Using Parrot, Chirp, and Hadoop

[slides]
Patrick Donnelly, Peter Bui and Douglas Thain
University of Notre Dame

Abstract: The Hadoop filesystem is a large scale distributed filesystem used to manage and quickly process extremely large data sets. We wish to utilize Hadoop to assist with data-intensive workloads in a distributed campus grid environment. Unfortunately, the Hadoop filesystem is not designed to work in such an environment easily or securely. We present a solution that bridges the Chirp distributed filesystem to Hadoop for simple access to large data sets. Chirp layers on top of Hadoop many grid computing desirables including simple deployment without special privileges, easy access via Parrot, and strong and flexible security ACLs. We discuss the challenges involved in using Hadoop on a campus grid and evaluate the performance of the combined systems.

elasticLM: A novel approach for software licensing in distributed computing infrastructures

[slides]
Wolfgang Ziegler, Claudio Cacciari, Francesco D'Andria, Bjorn Hagemeier, David Garcia Perez, Daniel Mallmann, Josep Martrat, Angela Rumpl, Csilla Zsigri and Miriam Gozalo
Fraunhofer Institute SCAI, CINECA, ATOS Origin, Forschungszentrum Julich GmbH, Foundation CESGA, Research Centre Juelich, JSC, ATOS Origin - Research and Innovation, The 451 Group

Abstract: A recent survey of the 451group on Cloud usage highlights software licensing as one of the top five obstacles for Cloud computing, quite similar to what has been observed in the Grid already a couple of years. The reasons are the same: the current praxis of software licensing, both in terms of business models and licensing technology. As a consequence, using commercial applications that require access to a license server for authorisation at run-time has been quite limited until recently in distributed computing environments, especially when the environment stretches across administrative domains like it is the case for public Clouds. In this paper we present a novel approach for managing software licenses as web service resources in distributed service oriented environments. Licenses become mobile objects, which may move to the environment where required to authorise the execution of a license protected application

A Novel Heuristic-based Task Selection and Allocation Framework in Dynamic Collaborative Cloud Service Platform

[slides]
Biao Song, Mohammad Mehedi Hassan and Eui-Nam Huh
Kyung Hee University, Global Campus

Abstract: To address interoperability and scalability issues for cloud computing, in our previous paper, we presented a novel cloud market model called CACM that enables a dynamic collaboration (DC) platform among different Cloud providers. As the initiator of dynamic collaboration, primary Cloud provider (pCP) needs an efficient local task selection and allocation algorithm to partition the whole tasks and allocate those tasks to be executed locally. Existing task allocation algorithms cannot be directly applicable in a DC environment since they may cause low resource utilization of local resources. So in this paper we propose a general task selection and allocation framework to improve resource utilization for pCP. The framework utilizes an adaptive filter to select tasks and a modified heuristic algorithm to allocate tasks. Moreover, a trade-off metric is developed as the optimization goal of heuristic algorithm, so that it is able to manage and optimize the trade-off between QoS of tasks and utilization of resources.

LEMO-MR: Low overhead and Elastic MapReduce Implementation Optimized for Memory and CPU-Intensive Applications

[slides]
Zacharia Fadika and Madhusudhan Govindaraju
Binghamton University, SUNY Binghamton

Abstract: Since its inception, MapReduce has frequently been associated with Hadoop and large-scale datasets. Its deployment at Amazon in the cloud, and its applications at Yahoo! and Face- book for large-scale distributed document indexing and database building, among other tasks, have thrust MapReduce to the fore- front of the data processing application domain. The applicability of the paradigm however extends far beyond its use with data intensive applications and diskbased systems, and can also be brought to bear in processing small but CPU intensive distributed applications. In this work, we focus both on the performance of processing large-scale hierarchical data in distributed scientific applications, as well as the processing of smaller but demanding input sizes primarily used in diskless, and memory resident I/O systems.

In this paper, we present LEMO-MR (Low overhead, elastic, configurable for in-memory applications, and on-demand fault tolerance), an optimized implementation of MapReduce, for both on disk and in memory applications, describe its architecture and identify not only the necessary components of this model, but also trade offs and factors to be considered. We show the efficacy of our implementation in terms of potential speedup that can be achieved for representative data sets used by cloud applications. Finally, we quantify the performance gains exhibited by our MapReduce implementation over Apache Hadoop in a compute intensive environment.

Exploring Architecture Options for a Federated, Cloud-based Systems Biology Knowledgebase

[slides]
Ian Gorton, Jenny Liu and Jian Yin
PNNL

Abstract: Systems biology is characterized by a large community of scientists who use a wide variety of fragmented and competing data sets and computational tools of all scales to support their research. In order to provide a more coherent computational environment for systems biology, we are working as part of the Department of Energy Systems Biology Knowledgebase (Kbase) project to define a federated cloud-based system architecture. The Kbase will eventually host massive amounts of biological data, provide high performance and scalable computational resources, and support a large user community with tools and services to enable them to utilize the Kbase resources. In this paper, we describe the results of our investigations into the design of a workflow infrastructure suitable for use in the Kbase. The approach utilizes standards-based workflow description and open source integration technologies, and incorporates a data aware workflow execution layer for exploiting data locality in the federated architecture. We describe a use case and the initial prototype implementation we have built that demonstrates the feasibility of our approach.

A Comparison and Critique of Eucalyptus, OpenNebula and Nimbus

[slides]
Peter Sempolinski and Douglas Thain
University of Notre Dame

Abstract: Eucalyptus, OpenNebula and Nimbus are three major open-source cloud-computing software platforms. The overall function of these systems is to manage the provisioning of virtual machines for a cloud providing infrastructure-as-a-service. These various open-source projects provide an important alternative for those who do not wish to use a commercially provided cloud.
In this paper, we provide a comparison of these systems and an analysis of each of these systems. We begin with a short summary comparing the raw feature set of these projects. After that we deepen of analysis by describing how these cloud management frameworks relate to the many other software components required to create a functioning cloud computing system. We also analyse the overall structure of each of these projects and address how the differing features and implementations reflect the different goals of each of these projects. Lastly, we discuss some of the common challenges that emerge in setting up any of these frameworks and suggest avenues of further research and improvement. These include the problem of fair scheduling in absence of money, eviction or preemption, the difficulties of network configuration, and the frequent lack of clean abstractions.

Image Distribution in Large Scale Cloud Providers

[slides]
Romain Wartel, Tony Cass, Belmiro Moreira, Ewan Roche, Manuel Guijarro, Sebastien Goasguen and Ulrich Schwickerath
CERN, Clemson University

Abstract: This paper presents the various mechanisms for virtual machine image distribution within a large batch farm and between sites that offer cloud computing services. The work is presented within the context of the Large Hadron Collider Computing Grid (LCG), it has two main goals. First it aims at presenting the CERN specific mechanisms that have been put in place to test the pre-staging of virtual machine images within a large cloud infrastructure of several hundred physical hosts. Second it introduces the basis of a policy for trusting and distributing virtual machine images between sites of the LCG. Finally experimental results are shown for the distribution of a 10 GB virtual machine image distributed to over 400 physical nodes using a binary tree and a BitTorrent algorithm. Results show that images can be pre-staged within 30 minutes.

Investigating Business-driven Cloudburst Schedulers for e-Science Bag-of-Tasks Applications

[slides]
David Maia, Ricardo Santos, Raquel Lopes and Francisco Brasileiro
Federal University of Campina Grande

Abstract: The new ways of doing science, rooted on the unprecedented processing, communication and storage infrastructures that became available to scientists, are collectively called e-Science. Many research labs now need non-trivial computational power during some periods of time, when they need to run e-science applications. Grid and voluntary computing are well-established solutions that cater to this need, but are not accessible for all labs and institutions. Besides, there is an uncertainty about the future amount of resources that will be available in such infrastructures, which prevents the researchers from planning their activities to guarantee that deadlines will be met.

With the emergence of the cloud computing paradigm come new opportunities. One possibility is to run e-science activities at resources acquired on-demand from cloud providers. However, although very low, there is a cost associated with the usage of cloud resources. Besides that, the amount of resources that can be simultaneously acquired is, in practise, limited. Another possibility is the not new idea of composing hybrid infrastructures in which the huge amount of computational resources shared by the grid infrastructures are used whenever possible and extra capacity is acquired from cloud computing providers. We here investigate how to schedule e-science activities in such hybrid infrastructures so that deadlines are met and costs are reduced. Our preliminary results indicate that online strategies can be used as an alternative to achieve better results than a strategy which prematurely acquires resources to run the application.

Evaluation and Analysis of GreenHDFS: A Self-Adaptive, Energy-Conserving Variant of the Hadoop Distributed File System

[slides]
Rini T Kaushik, Milind Bhandarkar and Klara Nahrstedt
Yahoo Inc., The University of Illinois Urbana-Champaign

Abstract: We present a detailed evaluation and sensitivity analysis of an energy-conserving, highly scalable variant of the Hadoop Distributed File System called GreenHDFS. Green-HDFS logically divides the servers in Hot and Cold Zones and relies on insightful data-classification driven energy-conserving data placement to realize guaranteed, substantially long periods (several days) of idleness in a significant subset of servers in the Hadoop cluster designated as the Cold Zone. Detailed lifespan analysis of the files in a large-scale Hadoop cluster at Yahoo! points at the viability of GreenHDFS. Simulation results with real-world traces from a Hadoop Cluster at Yahoo! show that GreenHDFS can achieve 24% energy cost reduction by doing power management in only one top-level tenant directory in the cluster. If GreenHDFS technique is applied to all the Hadoop clusters at Yahoo (amounting to 38000 servers), $2.1million can be saved in energy costs.

A Mechanism of Flexible Memory Exchange in Cloud Computing Environments

[slides]
Takeshi Okuda, Eiji Kawai and Suguru Yamaguchi
Nara Institute of Science and Technology Japan, National Institute of Information and Communications Technology Japan

Abstract: In cloud computing environment, virtual machine technology is used as a means of flexibly assigning workloads to real machines based on the static profiles of the workloads. Though virtual machine technology aims at flexible and dynamic resource sharing, memory is underutilized in cloud environment. We propose the virtual swap management mechanism (VSMM) which enables flexible and dynamic memory sharing through local area network. This VSMM can virtualize swap devices which a guest OS sees, and can switch underlying physical devices transparently to the guest OS. In this paper, we explain the architecture of the VSMM and give the detail of our prototype implementation using Xen open source hypervisor, open source iSCSI implementations, and logical volume management. Through various experiments, we demonstrate that VSMM can contribute to improve the performance of the processes and to make the most of equipped memory in cloud environments.

Clouds and Education

Correlation based File Prefetching Approach for Hadoop

[slides]
Bo Dong, Xiao Zhong, Qinghua Zheng, Lirong Jian, Jian Liu, Jie Qiu and Ying Li
Department of Computer Science and Technology, Xi'an Jiaotong University, IBM Research - China, Beijing, China

Abstract: Hadoop Distributed File System (HDFS) has been widely adopted to support Internet applications because of its reliable, scalable and low-cost storage capability. BlueSky, one of the most popular e-Learning resource sharing systems in China, is utilizing HDFS to store massive courseware. However, due to the inefficient access mechanism of HDFS, access latency of reading files from HDFS significantly impacts the performance of processing user requests. This paper introduces a two-level correlation based file prefetching approach, taking the characteristics of HDFS into consideration, to improve performance by reducing access latency. Four placement patterns to store prefetched data are presented, with policies to achieve trade-off between performance and efficiency of HDFS prefetching. Moreover, a dynamic replica selection algorithm is investigated to improve the efficiency of HDFS prefetching. The proposed correlation based file prefetching approach has been implemented in BlueSky, and experimental results prove that correlation based file prefetching can significantly reduce access latency therefore improve performance of Hadoop-based Internet applications.

Tree-Based Consistency Approach for Cloud Databases

[slides]
Susan Vrbsky and Md. Ashfakul Islam
University of Alabama

Abstract: Database as a Service is a very attractive product from cloud service providers for small start-up companies. However, consistency maintenance among replica servers is a very big issue in cloud databases. Too much interdependency of replica servers in the existing consistency model reduces performance and throughput, and can increase the transaction failure rate of a cloud database. In this paper we propose a tree-based consistency approach that reduces interdependency among replica servers by introducing partially consistent and fully consistent states of cloud databases. The tree is formed such a way that the maximum reliable path is ensured from the primary server to all replica servers. As a result, the probability of a transaction failure is greatly reduced, which helps to increase performance and throughput even in an unreliable network.

Efficient Metadata Generation to Enable Interactive Data Discovery over Large-scale Scientific Data Collections

[slides]
Sangmi Pallickara
Colorado State University

Abstract: Discovering the correct dataset efficiently is critical for computations and effective simulations in scientific experiments. In contrast to searching web documents over the Internet, massive binary datasets are difficult to browse or search. Users must select a reliable data publisher from the large collection of data services available over the Internet. Once a publisher is selected, the user must then discover the dataset that matches the computation’s needs, among tens of thousands of large data packages that are available. Some of the data hosting services provide advanced data search interfaces but their search scope is often limited to local datasets. Because scientific datasets are often encoded as binary data formats, querying or validating missing data over hundreds of Megabytes of a binary file involves a compute intensive decoding process. We have developed a system, GLEAN, that provides an efficient data discovery environment for users in scientific computing. Fine-grained metadata is automatically extracted to provide a micro view and profile of the large dataset to the users. We have used the Granules cloud runtime to orchestrate the MapReduce computations that extract metadata from the datasets. Here we focus on the overall architecture of the system and how it enables efficient data discovery. We applied our framework to a data discovery application in the atmospheric science domain. This paper includes a performance evaluation with observational datasets.

Reliability Support in Virtual Infrastructures

[slides]
Guilherme Koslovski, Wai-Leong Yeow, Cedric Westphal, Tram Truong Huu, Johan Montagnat and Pascale Vicat-Blanc
DoCoMo USA Labs, University of Nice - I3S, CNRS - I3S, INRIA - LYaTiss

Abstract: Through the recent emergence of joint resource and network virtualization, dynamic composition and provisioning of time-limited and isolated virtual infrastructures is now possible. One other benefit of infrastructure virtualization is the capability of transparent reliability provisioning (reliability becomes a service provided by the infrastructure). In this context, we discuss the motivations and gains of introducing customizable reliability of virtual infrastructures when executing large-scale distributed applications, and present a framework to specify, allocate and deploy virtualized infrastructure with reliability capabilities. An approach to efficiently specify and control the reliability at runtime is proposed. We illustrate these ideas by analyzing the introduction of reliability at the virtual-infrastructure level on a real application. Experimental results, obtained with an actual medical-imaging application running in virtual infrastructures provisioned in the experimental large-scale Grid'5000 platform, show the benefits of the virtualization of reliability.

Resource Provisioning for Enriched Services in Cloud Environment

[slides]
Rosy Aoun, Elias A. Doumith and Maurice Gagnaire
Telecom ParisTech and CNRS LTCI

Abstract: Cloud services are based on the provisioning of computing, storage, and networking resources in order to satisfy requests generated by remote end-users. High speed Internet access and multi-core Virtual Machines (VMs) enable today the provisioning of diversified and enriched types of services in Cloud environment. In this paper, we consider several types of basic services and show how their orchestration may lead to the provisioning of more sophisticated services. For this purpose, we define four types of requests that cover the wide spectrum of possible services. We then formulate the resource provisioning problem as a Mixed Integer Linear Program (MILP). We assume that the underlying infrastructure is based on a set of end-to-end connections with guaranteed sustainable bandwidth such as Carrier-Grade Ethernet (CGE) circuits. We investigate the impact of two innovative services on resource allocation carried out by Cloud Service Providers (CSPs).

These services correspond to distributed data storage and to multicast data transfer. For the former service, we consider the possibility of splitting a storage request onto different remote storage nodes. The latter service aims to distribute a same data sequence from one server towards multiple remote nodes assuming a limited number of network nodes has multicast capacities. These two innovative services provide a gain of 7% in terms of accepted requests when applied to the 18-node NSFnet backbone network.

Analyzing Electroencephalograms Using Cloud Computing Techniques

[slides]
Kathleen Ericson, Shrideep Pallickara and C. W. Anderson
Colorado State University

Abstract: Brain Computer Interfaces (BCIs) allow users to interact with a computer via electroencephalogram (EEG) signals generated by their brain. The BCI application that we consider allows a user to initiate actions such as keyboard input or control the motion of their wheelchair. Our goal is to be able to train the neural network and classify the EEG signals from multiple users to infer their intended actions in a distributed environment. The processing is developed using the Map-Reduce framework. We use our cloud runtime, Granules, to classify these EEG streams. One of our objectives is to be able to process these EEG streams in real-time. The BCI software has been developed in R, which is an interpreted language designed for the fast computation of matrix multiplications, making it an effective language for the development of artificial neural networks. We contrast our approach of using Granules with a competing approach that uses an R package – Snowfall that simplifies execution of R computations in a distributed setting. We have performed experiments to evaluate the costs introduced by our scheme for training the neural networks and classifying the EEG signals. Our results demonstrate the suitability of using Granules to classify multiple EEG streams in a distributed environment.

BetterLife 2.0: Large-scale Social Intelligence in Cloud Computing

[slides]
Dexter H. Hu, Yinfeng Wang and C.L. Wang
The University of Hong Kong

Abstract: This paper presents the design of the framework BetterLife 2.0 that implements large scale social intelligence application on the Cloud environment. We argued there is a growing number of applications needed to be implemented this way with a lot of user generated activities through online social websites. We adopted the Case-based Reasoning framework to provide logical reasoning. We outlined specific design considerations when porting a typical CBR framework jICOLIBRI2 on Cloud using Hadoop’s various services (e.g., MapReduce, HBases). These services allows efficient case base management (e.g. case insertion and adaptation) and computing intensive jobs distribution. With the scalability merit of MapReduce, we are able to provide recommendation service with social network analyzing for applications that can handle millions of users’ social activities.

Research Issues for Software Testing in the Cloud

[slides]
Leah Muthoni Riungu, Ossi Taipale and Kari Smolander
Lappeenranta University of Technology

Abstract: Cloud computing is causing a paradigm shift in the provision and use of computing services; away from the traditional desktop form to online services. This implies that the manner in which these computing services are tested should also change. This paper discusses the research issues that cloud computing imposes on software testing.  These issues were gathered from interviews with industry practitioners from eleven software organizations. The interviews were analyzed using qualitative grounded theory method. Findings of the study were compared with existing literature. The research issues were categorized according to application, management, legal and financial issues. The issues are discussed with the intention of soliciting academic research on software testing in the cloud. By addressing these issues, researchers can offer reliable recommendation for testing vendors and customers.

Computing Methods

CSAL: A Cloud Storage Abstraction Layer to Enable Portable Cloud Applications

[slides]
Zachary Hill and Marty Humphrey
University of Virginia

Abstract: One of the large impediments for adoption of cloud computing is perceived vendor lock-in with respect to both low-level resource management APIs and application-level storage services. Application portability is essential to both avoid lock-in as well as leverage the ever-changing landscape of cloud offerings. We present a storage API to enable applications to both utilize the highly-available and scalable storage services provided by cloud vendors and allow applications to be portable across platforms. Our API, called CSAL, provides Blob, Table, and Queue abstractions across multiple providers and presents applications with an integrated namespace thereby relieving applications of having to manage storage entity location information and access credentials. Overall, we have observed minimal overhead of CSAL on both EC2 and Windows Azure.

REMEM: REmote MEMory as Checkpointing Storage

[slides]
Hui Jin, Xian-He Sun, Yong Chen and Tao Ke
Illinois Institute of Technology, Oak Ridge National Laboratory

Abstract: Checkpointing is a widely used mechanism for supporting fault tolerance, but notorious in its high-cost disk access. The idea of memory-based checkpointing has been extensively studied in research but made little success in practice due to its complexity and potential reliability issues. In this study we present the design and implementation of REMEM, a REmote MEMory checkpointing system to extend the checkpointing storage from disk to remote memory. A unique feature of REMEM is that it can be integrated into existing disk-based checkpointing systems seamlessly. A user can flexibly switch between REMEM and disk as checkpointing storage to balance the efficiency and reliability. The implementation of REMEM on Open MPI is also introduced. The experimental results confirm that REMEM and the proposed adaptive checkpointing storage selection are promising in both performance, reliability and scalability.

Performance Analysis of High Performance Computing Applications on the Amazon Web Services Cloud

[slides]
Keith Jackson, Lavanya Ramakrishnan, Krishna Muriki, Shane Canon, Shreyas Cholia, John Shalf, Harvey Wasserman and Nicholas Wright
Microsoft, LBNL

Abstract: Cloud computing has seen tremendous growth, particularly for commercial web applications. The on-demand, pay-as-you-go model creates a flexible and cost-effective means to access compute resources. For these reasons, the scientific computing community has shown increasing interest in exploring cloud computing. However, the underlying implementation and performance of clouds are very different from those at traditional supercomputing centers. It is therefore critical to evaluate the performance of HPC applications in today’s cloud environments to understand the tradeoffs inherent in migrating to the cloud. This work represents the most comprehensive evaluation to date comparing conventional HPC platforms to Amazon EC2, using real applications representative of the workload at a typical supercomputing center. Overall results indicate that EC2 is six times slower than a typical mid-range Linux cluster, and twenty times slower than a modern HPC system. The interconnect on the EC2 cloud platform severely limits performance and causes significant variability.

Exploring the Performance Fluctuations of HPC Workloads on Clouds

[slides]
Yaakoub El Khamra, Hyunjoo Kim, Shantenu Jha and Manish Parashar
Texas Advanced Computing Center, Rutgers University, CCT / LSU

Abstract: Clouds can support novel execution modes often supported by advanced capabilities such as autonomic schedulers. These capabilities are predicated upon an accurate estimation and calculation of runtimes on a given infrastructure. Using a well understood HPC workload, we find strong fluctuations from the mean performance on EC2 and Eucalyptus-based cloud systems. Our analysis eliminates variations in IO and computational times as possible causes; we find that variations in communication times account for the bulk of the experiment-to-experiment fluctuations of the performance.

Finding Tropical Cyclones on a Cloud Computing Cluster: Using Parallel Virtualization for Large-Scale Climate Simulation Analysis

[slides]
Daren Hasenkamp, Alex Sim, Michael Wehner and Kesheng Wu
Lawrence Berkeley National Laboratory

Abstract: Extensive computing power has been used to tackle issues such as climate changes, fusion energy, and other pressing scientific challenges. In this work, we bring the power of cloud computing to bear on the task of analyzing trends of tropical cyclones in climate simulation data. The cloud computing platform is attractive here because it can provide an environment familiar to climatologists and their analysis tools. We created virtual machines (VMs) and ran them on 2 different cloud clusters. Our VM communicates either with instances of itself or with a remote server to split up and analyze large datasets in parallel. In a preliminary test, we used this virtual climate analysis platform to analyze ~500GB of climate data. Using 30 VMs, the total analysis time was reduced by a factor of ~21 from the traditional, personal workstation-based analysis method. The main advantages of our method are that the level of parallelism is easily configurable, and software dependency resolution is simple. This initial work demonstrates that a cloud computing system is a viable platform for distributed scientific data analysis traditionally conducted on dedicated supercomputing systems.

Evaluation of MapReduce for Gridding LIDAR Data

[slides]
Sriram Krishnan, Christopher Crosby and Chaitanya Baru
San Diego Supercomputer Center

Abstract: The MapReduce programming model, introduced by Google, has become popular over the past few years as a mechanism for processing large amounts of data, using shared-nothing parallelism. In this paper, we investigate the use of MapReduce technology for a local gridding algorithm for the generation of Digital Elevation Models (DEM). The local gridding algorithm utilizes the elevation information from LIDAR (Light, Detection, and Ranging) measurements contained within a circular search area to compute the elevation of each grid cell. The method is data parallel, lending itself to implementation using the MapReduce model. Here, we compare our initial C++ implementation of the gridding algorithm to a MapReduce-based implementation, and present observations on the performance (in particular, price/performance) and the implementation complexity. We also discuss the applicability of MapReduce technologies for related applications.

Cost-effective HPC: The Community or the Cloud?

[slides]
Adam Carlyle, Stephen Harrell and Preston Smith
Purdue University

Abstract: The increasing availability of commercial cloud computing resources in recent years has caught the attention of the high-performance computing (HPC) and scientific computing community. Many researchers have subsequently examined the relative computational performance of commercially available cloud computing offerings across a number of HPC application benchmarks and scientific workflows, but the analogous cost comparisons i.e., comparisons between the cost of doing scientific computation in traditional HPC environments vs. cloud computing environments - are less frequently discussed and are difficult to make in meaningful ways. Such comparisons are of interest to traditional HPC resource providers as well as to members of the scientific research community who need access to HPC resources on a routine basis.

This paper is a case study of costs incurred by faculty end-users of Purdue University's HPC ``community cluster'' program. We develop and present a per node-hour cloud computing equivalent cost that is based upon actual usage patterns of the community cluster participants and is suitable for direct comparison to hourly costs charged by one commercial cloud computing provider. We find that the majority of community cluster participants incur substantially lower out-of-pocket costs in this community cluster program than in purchasing cloud computing HPC products. In addition, we find that the overall costs incurred by the institution in providing HPC resources and support to the research community is also lower than comparative commercial cloud costs. We also characterize the types of usage scenarios where cloud computing can be financially beneficial to members of the scientific computing community.

Recommendations for Virtualization Technologies in High Performance Computing

[slides]
Nathan Regola and Jean-Christophe Ducom
University of Notre Dame

Abstract: The benefits of virtualization are typically considered to be server consolidation, (leading to the reduction of power and cooling costs) increased availability, isolation, ease of operating system deployment and simplified disaster recovery. High Performance Computing (HPC) environments pose one main challenge for virtualization: the need to maximize throughput with minimal loss of CPU and I/O efficiency. However, virtualization is usually evaluated in terms of enterprise workloads and assumes that servers are underutilized and can be consolidated. In this paper we evaluate the performance of several virtual machine technologies in the context of HPC. A fundamental requirement of current high performance workloads is that both CPU and I/O must be highly efficient for tasks such as MPI jobs.

This work benchmarks two virtual machine monitors, OpenVZ and KVM, specifically focusing on I/O throughput since CPU efficiency has been extensively studied [1]. OpenVZ offers near native I/O performance. Amazon’s EC2 “Cluster Compute Node” product is also considered for comparative purposes and performs quite well. The EC2 “Cluster Compute Node” product utilizes the Xen hypervisor in hvm mode and 10 Gbit/s Ethernet for high throughput communication. Therefore, we also briefly studied Xen on our hardware platform (in hvm mode) to determine if there are still areas of improvement in KVM that allow EC2 to outperform KVM (with InfiniBand host channel adapters operating at 20 Gbit/s) in MPI benchmarks. We conclude that KVM’s I/O performance is suboptimal, potentially due to memory management problems in the hypervisor. Amazon’s EC2 service is promising, although further investigation is necessary to understand the effects of network based storage on I/O throughput in compute nodes. Amazon’s offering may be attractive for users searching for “InfiniBand-like” performance without the upfront investment required to build an InfiniBand cluster or users wishing to dynamically expand their cluster during periods of high demand.

Cost-Optimal Outsourcing of Applications into the Clouds

[slides]
Immanuel Trummer, Frank Leymann, Ralph Mietzner and Walter Binder
Artificial Intelligence Laboratory, Ecole Polytechnique Federale de Lausanne, Institute of Architecture of Application Systems, University of Stuttgart Faculty of Informatics, University of Lugano

Abstract: Commercial services for provisioning software components and virtual infrastructure in the cloud are emerging. For customers, this creates a multitude of possibilities for outsourcing part of the IT-stack to third parties in order to run their applications. These possibilities are associated with different running costs, so cloud customers have to determine the optimal solution. In this paper, we present and experimentally evaluate an algorithm that solves the corresponding optimization problem.

We assume that applications are described as templates, fixing the deployment structure and constraining the properties of the used soft- and hardware components. Different parts of the application may be outsourced to different providers and several levels of outsourcing can be considered. However, dependencies between different parts of the application have to be respected. Our algorithm decomposes the application graph in a first step in order to discover all suitable cloud provisioning services from a registry. It determines the optimal solution by representing the problem as constraint optimization problem that can be solved by an existing solver implementation.

A Novel Approach for Cooperative Overlay-Maintenance in Multi-Overlay Environments

[slides]
Chin-Jung Hsu, Wu-Chun Chung, Kuan-Chou Lai, Kuan-Ching Li and Yeh-Ching Chung
National Tsing Hua University, National Taichung University, Providence University

Abstract: Overlay networks are widely adopted in many distributed systems for efficient resource sharing: recently, the overlay network is also introduced into the cloud system to organize thousands of virtualized resources. The explosion of P2P applications introduces the multi-overlay environment in which a number of nodes simultaneously participate in multiple overlays. When there are multiple applications running over a large set of nodes, some nodes may take repeated efforts to preserve multi-overlay networks. Therefore, maintaining these co-existing overlays brings the redundant maintenance overhead. This paper presents a cooperative strategy to analyze the overlay maintenance of a multi-overlay environment and to elaborate multiple overlays for simplifying the overlay maintenance. The proposed strategy exploits the synergy of co-existing overlays to handle their common overlay-maintenance, so that the redundant maintenance overhead could be eliminated while keeping the performance. To evaluate the system performance, this paper not only analyzes several overlays but also considers more realistic multi-overlay environments by varying the intersection ratio of diverse overlays and the combination of multiple overlays. Experimental results show that the proposed cooperative strategy significantly decreases the redundant overlay-maintenance overhead, where the reduction ratio of maintaining multiple overlays is higher than 60 percent in some cases.

Combinatorial Auction-Based Allocation of Virtual Machine Instances in Clouds

[slides]
Sharrukh Zaman and Daniel Grosu
Wayne State University

Abstract: The current cloud computing platforms allocate virtual machine instances to their users through fixed-price allocation mechanisms. We argue that combinatorial auction-based allocation mechanisms are especially efficient over the fixed-price mechanisms since the virtual machine instances are assigned to users having the highest valuation. We formulate the problem of virtual machine allocation in clouds as a combinatorial auction problem and propose two mechanisms to solve it. We perform extensive simulation experiments to compare the two proposed combinatorial auction-based mechanisms with the currently used fixed-price allocation mechanism. Our experiments reveal that the combinatorial auction-based mechanisms can significantly improve the allocation efficiency while generating higher revenue for the cloud providers.

A Multi-agent approach for Semantic Resource Allocation

[slides]
Jorge Ejarque, Raul Sirvent and Rosa M. Badia Barcelona Supercomputing Center

Abstract: This paper presents an new approach of the Semanti- cally Enhanced Resource Allocation (SERA) distributed as a multi-agent system. It presents a distributed resource al- location process which combines the benefits of semantic web for making easier the integration between multiple resource providers in the Cloud and agent technologies for coordinating and adapting the execution accross the different providers. The allocation process is based on the nego- tiation of different agents which allows the combination of customer and providers policies getting scheduling results which satifies both parts. The SERA agents can be deployed in multiple locations improving the system scalability. The new approach makes the SERA suitable for working as a scheduler inside a Service Provider as well as a metascheduler integrating resources from different providers and platforms (clusters, grids, clouds,...).

Forecasting for Grid and Cloud Computing On-Demand Resources Based on Pattern Matching

[slides]
Frederic Desprez, Eddy Caron and Adrian Muresan
ENS-Lyon / INRIA

Abstract: The Cloud phenomenon brings along the cost-saving benefit of dynamic scaling. As a result, the question of efficient resource scaling arises. Prediction is necessary as the virtual resources that Cloud computing uses have a setup time that is not negligible. We propose a new approach to the problem of workload prediction based on identifying similar past occurrences of the current short-term workload history.

We present in detail the Cloud client resource auto-scaling algorithm that uses the above approach to help when scaling decisions are made, as well as experimental results by using real-world traces from Cloud and grid platforms. We also present an overall evaluation of this approach, its potential and usefulness for enabling efficient auto-scaling of Cloud user resources.

Self-Caring IT Systems: A Proof-of-Concept Implementation in Virtualized Environments

[slides]
Selvi Kadirvel and Jose A.B. Fortes
University of Florida

Abstract: In self-caring IT systems, faults are handled proactively, e.g. by slowing down the deterioration of system health thereby effectively avoiding or delaying system failures. This requires health management which entails health monitoring, diagnosis, prognosis, planning of recovery and remediation actions. A brief overview of our prior work, which proposes a general methodology to capture system properties and incorporate health management using Petri nets, is provided. We describe in detail an application of the proposed formal method to the design and development of middleware that can manage the health of a batch-based, job submission system on a virtualized platform.

First, we describe how a real world job submission IT system is converted to a Petri net model. Secondly, we show system validation and analysis using this model to understand resource needs of different activities in the IT chain. Thirdly, we describe how the executable model is used as a system manager to control operation and health management of a virtualized test bed. Fourthly, we illustrate the use of a feedback controller to manage health deterioration due to resource depletion in the job-execution stage of the modeled IT chain.

Using a proof-of-concept implementation, we show that the early detection and handling of health deteriorations results in significant benefits in terms of cost savings and down time reduction. Experimental results show that our health management framework can be used to effectively prevent job failures, while imposing low overhead to the managed system. We have shown that for a typical workload consisting of jobs that suffer from potential resource depletion faults, our feedback controller can be used to gain useful life that is needed for critical planning and remediation actions in up to 82% of the jobs..

MapReduce

CloudBATCH: A Batch Job Queuing System on Clouds with Hadoop and HBase

[slides]
Chen Zhang and Hans De Sterck
University of Waterloo

Abstract: As MapReduce becomes more and more popular in data processing applications, the demand for Hadoop cluster grows increasingly. However, Hadoop is normally incompatible with existing cluster batch job queuing systems and requires a dedicated cluster under its full control. Hadoop also lacks supports for user access control, accounting, fine-grain performance monitoring and legacy batch job processing facilities comparable to existing cluster job queuing systems, making dedicated Hadoop clusters less amenable for administrators and normal users alike with hybrid computing needs involving both MapReduce and legacy applications. As a result, getting a properly suited and sizable Hadoop cluster has not been easy in organizations with existing clusters. This paper presents CloudBATCH, a prototype solution to this problem by enabling Hadoop to function as a traditional batch job queuing system with enhanced management functionality for cluster resource management. With CloudBATCH, a complete shift to Hadoop for managing an entire cluster to cater for hybrid computing needs becomes feasible. It also makes dedicated Hadoop clusters useful for the load balancing of legacy batch job submissions to other clusters managed by traditional queuing systems.

Voronoi-based Geospatial Query Processing with MapReduce

[slides]
Afsin Akdogan, Ugur Demiryurek, Farnoush Banaei-Kashani and Cyrus Shahabi
University of Southern California

Abstract: Geospatial queries (GQ) have been used in a wide variety of applications such as decision support systems, profile-based marketing, bioinformatics and GIS. Most of the existing query-answering approaches assume non parallel processing on a single machine although GQs are intrinsically parallelizable. There are some approaches that have been designed for parallel databases and cluster systems; however, these only apply to the systems with limited parallel processing capability, far from that of cloud-based platforms. In this paper, we study the problem of parallel geospatial query processing with MapReduce programming model. Our proposed approach first creates a spatial index, Voronoi diagram, for given data points in 2D space and enables efficient processing of a wide range of GQs. We evaluated the performance of our proposed techniques and correspondingly compared them with their closest related work while varying the number of employed nodes.

MapReduce in the Clouds for Science

[slides]
Thilina Gunarathne, Tak-lon Wu, Judy Qiu and Geoffrey C. Fox
Indiana University

Abstract: Utility computing model introduced by cloud computing together with the rich set of cloud infrastructure services offers a very viable alternative for the traditional servers and compute clusters. MapReduce distributed data processing architecture has become the weapon of choice for data intensive analyses in the clouds and in commodity clusters due to its fault tolerance features, scalability and the ease of use. Currently there are several options for using MapReduce in the cloud environments such as using MapReduce as a service, setting up your own MapReduce cluster on cloud instances as well as using specialized cloud MapReduce runtimes which take advantage of the cloud infrastructure services. In this paper we evaluate the use and performance of MapReduce in the cloud environments for scientific applications using DNA sequence assembly and sequence alignment as use cases. We also introduce and evaluate the concept of AzureMapReduce, a novel MapReduce runtime build using the Microsoft Azure cloud infrastructure services.

Scheduling Hadoop Jobs to Meet Deadline Constraints

[slides]
Kamal KC and Kemafor Anyanwu
North Carolina State University

Abstract: User constraints such as deadlines are important requirements that are not considered by existing cloud-based data processing environments such as Hadoop. In the current implementation, jobs are scheduled in FIFO order by default with options for other priority based schedulers. In this paper, we extend an approach for real time cluster scheduling approach to account for the two-phase computation style of MapReduce. We develop criteria for scheduling jobs based on user specified deadline constraints and discuss our implementation and preliminary evaluation of a Deadline Constraint Scheduler for Hadoop that ensures that only jobs whose deadlines can be met are scheduled for execution.

Data Acquisition in Hadoop System

[slides]
Baodong Jia, Tomasz Wiktor Wlodarczyk and Chunming Rong
Stavanger University, Norway

Abstract: Data has become more and more important these years, especially for big companies, and it is of great benefit to dig out useful information inside. In Oil & Gas industry, there are a lot of data available, both in real-time and historical format. As the amount of data is huge, it is usually infeasibleor very time consuming to process the data. Hadoop is introduced to solve this problem. In order to perform Hadoop jobs, data must exist on the Hadoop file system, which brings the problem of data acquisition. In this paper, two solutions are given out for data acquisition. The performance comparison is introduced afterwards, and solution based on Chukwa is proved to be bet-ter than the other one.

Security and Risk

Trusted Data Sharing over Untrusted Cloud Storage Providers

[slides]
Gansen Zhao, Chunming Rong and Jin Li
University of Stavanger, Norway

Abstract: Cloud computing has been acknowledged as one of the prevaling models for providing IT capacities. The off-premises computing paradigm that comes with cloud computing has incurred great concerns on the security of data, especially the integrity and confidentiality of data, as cloud service providers may have complete control on the computing infrastructure that underpins the services. This makes it difficult to share data via cloud providers where data should be confidential to the providers and only authorized users should be allowed to access the data.

This work aims to construct a system for trusted data sharing throught untrusted cloud providers to address the above mentioned issue. The constructed system can imperatively impose the access control policies of data owners, preventing the cloud storage providers from unauthorized access and making illegal authorization to access the data.

SafeVanish: An Improved Data Self-Destruction for Protecting Data Privacy

[slides]
Lingfang Zeng, Zhan Shi, Shengjie Xu and Dan Feng
Huazhong University of Science and Technology,

Abstract: In the background of cloud, self-destructing data mainly aims at protecting the data privacy. All the data and its copies will become destructed or unreadable after a user-specified period, without any user intervention. Besides, anyone cannot get the decryption key after timeout, neither the sender nor the receiver. The Washington’s Vanish system is a system for self-destructing data under cloud computing, and it is vulnerable to "hopping attack" and "sniffer attack". We propose a new scheme in this paper, called SafeVanish, to prevent hopping attacks by way of extending the length range of the key shares to increase the attack cost substantially, and do some improvement on the Shamir Secret Sharing algorithm implemented in the Original Vanish system. We present an improved approach against sniffing attacks by using the public key cryptosystem to protectt from sniffing operations. In addition, we evaluate analytically the functionality of the proposed SafeVanish system.

Intercloud Security Considerations

[slides]
David Bernstein and Deepak Vij
Huawei Technologies, Cloud Strategy Partners

Abstract: Cloud computing is a new design pattern for large, distributed datacenters. Service providers offering applications including search, email, and social networks have pioneered this specific to their application. Recently they have expanded offerings to include compute-related capabilities such as virtual machines, storage, and complete operating system services. The cloud computing design yields breakthroughs in geographical distribution, resource utilization efficiency, and infrastructure automation. These “public clouds” have been replicated by IT vendors for corporations to build “private clouds” of their own. Public and private clouds offer their end consumers a “pay as you go” model - a powerful shift for computing, towards a utility model like the electricity system, the telephone system, or more recently the Internet. However, unlike those utilities, clouds cannot yet federate and interoperate. Such federation is called the “Intercloud”. Building the Intercloud is more than technical protocols. A blueprint for an Intercloud economy must be architected with a technically sound foundation and topology. As part of the overall Intercloud Topology, this paper builds on the technology foundation emerging for the Intercloud and specifically delves into details of Intercloud security considerations such as Trust Model, Identity and Access Management, governance considerations and so on.

Application-Oriented Remote Verification Trust Model in Cloud Computing

[slides]
Xiaofei Zhang, Hui Liu, Xing Wang and Shizhong Wu
China Information Technology Security Evaluation Center

Abstract: The emergence and application of cloud computing can help users access to various computing resources and services more conveniently. However, it also brings forth many security challenges. This paper proposes the application-oriented remote verification trust model, which is capable of adjusting the user’s trust authorization verification contents according to the specific security requirements of different applications. The model also dynamically adjusts the users’ trust value with the trust feedback mechanism to determine whether or not the requested resource or service should be provided, so as to guarantee the security of information resources. This paper provides a formal description of the basic components and trust properties of the model with a belief formula, and describes the framework for the implementation of the model.

Visualization

Building a Distributed Block Storage System for Cloud Infrastructure

[slides]
Xiaoming Gao, Yu Ma, Marlon Pierce, Mike Lowe and Geoffrey Fox
Pervasive Technology Institute, Indiana University

Abstract: The development of cloud infrastructures has stimulated interest in virtualized block storage systems, exemplified by Amazon Elastic Block Store (EBS), Eucalyptus’ EBS implementation, and the Virtual Block Store (VBS) system. Compared with other solutions, VBS is designed for flexibility, and can be extended to support various Virtual Machine Managers and Cloud platforms. However, due to its single-volume-server architecture, VBS has the problem of single point of failure and low scalability. This paper presents our latest improvements to VBS for solving these problems, including a new distributed architecture based on the Lustre file system, new workflows, better reliability and scalability, and read-only volume sharing. We call this improved implementation VBS-Lustre. Performance tests show that VBS-Lustre can provide both better throughput and higher scalability in multiple-attachment scenarios than VBS. VBS-Lustre could potentially be applied to solve some challenges for current cluster file systems, such as metadata management and small file access.

Dynamic Resource Provisioning for Data Streaming Applications in a Cloud Environment

[slides]
Smita Vijayakumar, Qian Zhu and Gagan Agrawal
The Ohio State University

Abstract: The recent emergence of cloud computing is making the vision of utility computing realizable, i.e., computing resources and services from a cloud can be delivered, utilized, and paid for in the same fashion as utilities like water or electricity. Current cloud service providers have taken some steps towards supporting the true {\em pay-as-you-go} or a utility-like pricing model, a
nd current research points towards more fine-grained allocation and pricing of resources in the future. In such environments, resource provisioning becomes a challenging problem, since one needs to avoid both under-provisioning (leading to application slowdown) and over-provisioning (leading to unnecessary resource costs). In this paper, we consider this problem in the context of streaming applications. In these applications, since the data is generated by external sources, the goal is to carefully allocate resources so that the processing rate can match the rate of data arrival. We have developed a solution that can handle unexpected data rates, including the transient rates. We evaluate our approach using two streaming applications in a virtualized environment.

Affinity-aware Dynamic Pinning Scheduling for Virtual Machines

[slides]
Zhi Li, Yuebin Bai, Huiyong Zhang and Yao Ma
School of Computer Science, Beihang University

Abstract: Virtualization provides an effective management in server consolidation. The transparence enables different kinds of servers running in the same platform, making full use of hardware resource. However, virtualization introduces two-level schedulers: one from Guest OS, where the tasks are scheduled to virtual CPUs (VCPUs); the other from the virtual machine monitor (VMM), where VCPUs are scheduled to CPUs. As a result, the lower level scheduler is ignorant of the task information so that it cannot allocate appropriate proportion of CPU resource for every Guest OS in some cases. This paper presents an affinity-aware Dynamic Pinning Scheduling scheduler (DP-Scheduling). We aim at two objects: Bridging the semantic gap between Guest OS and VMM, introducing an affinity-aware method and providing the tasks information about CPU affinity to VMM; Bringing up a novel scheduling, DP-Scheduling, so that VCPU can be pinned or unpinned on one CPU’s running queue dynamically. For this purpose, we first get the Machine Address (MA) of process descriptor from the angle of VMM. The affinity information is also acquired before the task is enabled to run. To acknowledge the affinity information, DP-Scheduling calls an API provided by us. Depending on the affinity information, we put forward a series of measures to implement pinning dynamically as well as to keep workload balance. All implementation is confined to Xen VMM and Credit scheduler. Our experiments demonstrate that DP-Scheduling outperforms Credit scheduling by testing various indicators for CPU-bound tasks, without interfering the load balance.

Achieving High Throughput by Transparent Network Interface Virtualization on Multi-core Systems

[slides]
Huiyong Zhang, Yuebin Bai, Zhi Li, Niandong Du and Wentao Yang
Beihang University

Abstract: Though with the rapid development, there remains a challenge on achieving high performance of I/O virtualization. The paravirtualized I/O driver domain model, used in Xen, provides several advantages including fault isolation, live migration, and hardware independence. However, the high CPU overhead of driver domain leads to low throughput for high bandwidth links. Direct I/O can achieve high performance but at the cost of removing the benefits of the driver domain model. This paper presents software techniques and optimizations to achieve high throughput network I/O virtualization by driver domain virtualization model on multi-core systems. In our experiments on multi-core system with a quad-port 1GbE NIC, we observe the overall throughput of multiple guest VMs can only be 2.2Gb/s, while the link bandwidth is 4Gb/s in total. The low performance results from the disability of driver domain to concurrently serve multiple guest VMs running bandwidth-intensive applications. Consequently, two approaches are proposed. First, a multi-tasklet netback is implemented to serve multiple netfronts concurrently. Second, we implement a new event channel dispatch mechanism to balance event associated with network I/O over VCPUs of driver domain. To reduce the CPU overhead of the driver domain model, we also propose two optimizations: lower down event frequency in netback and implement LRO in netfront. By applying all the above techniques, our experiments show that the overall throughput can be improved from the original 2.2Gb/s to 3.7Gb/s and the multi-core CPU resources can be utilized efficiently. We believe that the approaches of our study can be valuable for high throughput I/O virtualization in the coming multi-core era..

Xenrelay: An Efficient Data Transmitting Approach for Tracing Guest Domain

[slides]
Hai Jin, Wenzhi Cao, Pingpeng Yuan and Xia Xie
Huazhong University of Science and Technology

Abstract: As the degree of virtualization is growing considerably, improving performance of virtual machine environments motivates deeper investigation of the internal processes and performance implications of virtualization. Several tools are currently available to help analyze performance in virtual machine environments. However, all of these cannot support users to add their own trace information.

In this paper, we have presented Xenrelay, a unified, efficient, and simple mechanism for transferring large amounts of data between the guest domain kernel and the privileged domain user-space. Xenrelay allows users (who trace subsystems of guest domain to analyze performance) to record and fast relay data between domains.

We provide an example of Xenrelay’s value in revealing behavior of virtualization. We build a block trace toolkit (use Xenrelay as an engine). Then, we use this toolkit to track the block I/O virtualization procedures and evaluated overheads in the Xen environment. Our trace results identify that the main delay is the time of request transmission and block device operation. Besides, the results reveal impact of the I/O schedules on the disk performance.

Miscellaneous

Petri Net Modeling of the Reconfigurable Protocol Stack for Cloud Computing based Control Systems

[slides]
Hui Chen, Chunjie Zhou and Naixue Xiong
Huazhong University of Science and Technology, Georgia State University

Abstract: The Industrial Ethernet is promising for the implementation of a Cloud Computing based control system. However, numerous standard organizations and vendors have developed various Industrial Ethernets to satisfy the real-time requirements of field devices. This paper presents a real-time reconfigurable protocol stack to cope with this challenge, by introducing the architecture with a core of dynamic routing and autonomic local scheduling. It is based on the deterministic and stochastic Petri-Nets (DSPN) method to illustrate the performance of producer/consumer based application model, CSMA/CD based node accessing activities, and TDMA based resource allocation for real time and non-real time traffic. Furthermore, the predicted time distribution for evaluating the stability of a control system can be obtained from the proposed DSPN model. It is shown that the DSPN modeling yields good verification analysis and performance prediction results through a real experimentation.

A Token-Based Access Control System for RDF Data in the Clouds

[slides]
Arindam Khaled, Mohammad Husain, Latifur Khan, Kevin Hamlen and Bhavani Thuraisingham
Mississippi State University, University of Texas at Dallas

Abstract: Semantic Web is gaining immense popularity. Resource Description Framework (RDF) is broadly used for Semantic Web. Apart from Jena, access control on other RDF stores used for single machines has been discussed seldom in the literature. One significant obstacle to using RDF stores defined for single machines is their scalability. Cloud computers, on the other hand, have been proven useful for storing large RDF stores. But these systems lack access control on RDF data as to our knowledge. We propose a token based access control system that is being implemented in Hadoop (an open source cloud computing framework). We define six types of access levels. The access control policies have been implemented in three levels -- Query Rewriting, Embedded Enforcement, and Post-processing Enforcement. In Embedded Enforcement, policies are enforced during data selection using MapReduce where as in Post-processing Enforcement they are enforced during the presentation of data to the users. Experiments show that  Embedded Enforcement outperforms Post-processing Enforcement.

Data replication and power consumption in data grids

[slides]
Susan Vrbsky, Ming Lei, Karl Smith and S. Jeff Byrd
University of Alabama

Abstract: While data grids can provide the ability to solve large-scale applications which require the processing of large amounts of data, they have been recognized as extremely energy inefficient.  Computing elements can be located far away from the data storage elements.  A common solution to improve availability and file access time in such environments is to replicate the data, resulting in the creation of copies of data files at many different sites.  The energy efficiency of the data centers storing this data is one of the biggest issues in data intensive computing.  Since power is needed to transmit, store and cool the data, we propose to minimize the amount of data transmitted and stored by utilizing smart replication strategies that are data aware.   In this paper we present a new data replication approach, called the sliding window replica strategy (SWIN), that is not only data aware, but is also energy efficient.  We measure the performance of SWIN and existing replica strategies on our Sage green cluster to study the power consumption of the strategies.  Results from this study have implications beyond our cluster to the management of data in clouds.

User Demand Prediction from Application Usage Pattern in Virtual Smartphone

[slides]
Joon Heo, Kenji Terada, Masashi Toyama, Shunsuke Kurumatani and Eric Y. Chen
NTT Corporation

Abstract: The numbers of smartphone users and related applications are growing rapidly, and applications continue to become more data-intensive. In the cloud based service for smartphone, if user demand on virtual machines exceeds the hardware capacity of the server, the server incurs an overload and bottleneck; network delay, latency, and packet loss rate are increased in 3G and Wi-Fi connections. Therefore, it is important to predict user demand and to use this information for resource allocation methods such as network virtualization and load balancing. We present a novel user demand prediction method that uses analysis results of application usage patterns. By analysis of log data and using the proposed method, we can predict execution time and average volume of transmitted application data. The proposed method is mainly considered for adoption in our virtual smartphone system. We show results from an experiment performed in an implemented test-bed, including prediction results and performance of wireless media.

Semantics Centric Solutions for Application and Data Portability in Cloud Computing

[slides]
Ajith Ranabahu and Amit Sheth
Knoesis Center, Wright State University

Abstract: Cloud computing has become one of the key considerations both in academia and industry. Cheap, seemingly unlimited computing resources that can be allocated almost instantaneously and pay-as-you-go pricing schemes are some of the reasons for the success of Cloud computing. The Cloud computing landscape however is plagued by many issues hindering adoption. One such issue is vendor lock-in, forcing the Cloud users to adhere to one service provider in terms of data and application logic. Semantic Web has been an important research area that has seen significant attention from both academic and industrial researchers. One key property of Semantic Web is the notion of interoperability and portability through high level models. Significant work has been done in the areas of data modeling, matching and transformations. The issues the Cloud computing community is facing now with respect to portability of data and application logic are exactly the same issue the Semantic Web community has been trying to address for some time. In this paper we present an outline of the use of well established semantic technologies to overcome the vendor lock-in issues in Cloud computing. We present a semantics-centric programming paradigm to create portable Cloud applications and discuss MobiCloud, our early attempt to implement the proposed approach.

Power of Clouds In Your Pocket: An Efficient Approach for Cloud Mobile Hybrid Application Development

[slides]
Ashwin Manjunatha, Ajith Ranabahu, Amit P. Sheth and Krishnaprasad Thirunarayan
Knoesis Center, Wright State University

Abstract: The advancements in computing have resulted in a boom in cheap, ubiquitous, connected mobile devices as well as seemingly unlimited, utility style, pay as you go computing resources, commonly referred to as Cloud computing. However, taking full advantage of this mobile and cloud computing landscape, especially for the data intensive domains has been hampered by the many heterogeneities that exist in the mobile space as well as the Cloud space. Our research focuses on exploiting the capabilities of the mobile and cloud landscape by defining a new class of applications called cloud mobile hybrid (CMH) Applications and a Domain Specific Language (DSL) based methodology to develop these applications. We define Cloud-mobile hybrid as a collective application that has a Cloud based back-end and a mobile device front-end. Using a single DSL script, our current toolkit is capable of generating a variety of CMH applications. These applications are composed of multiple combinations of native Cloud and mobile applications. Our approach not only reduces the learning curve but also shields the developers from the complexities of the target platforms. We provide a detailed description of our language and present the results obtained using our prototype generator. We also present a list of extensions we are working on related to this research.

Rapid Processing of Synthetic Seismograms Using Windows Azure Cloud

[slides]
Vedaprakash Subramanian and Liqiang Wang
University of Wyoming

Abstract: Currently, numerically simulated synthetic seismograms are widely used by seismologists for seismological inferences. The generation of these synthetic seismograms requires large amount of computing resources, and the maintenance of these observed seismograms requires massive storage. Traditional high-performance computing platforms is inefficient to handle these applications because rapid computations are needed and large-scale datasets should be maintained. The emerging cloud computing platform provides an efficient substitute. In this paper, we introduce our experience on implementing a computational platform for rapidly computing and delivering synthetic seismograms on Windows Azure. Our experiment shows that cloud computing is an ideal platform for such kind of applications.

Power-Saving in Large-Scale Storage Systems with Data Migration

[slides]
Koji Hasebe, Tatsuya Niwa, Akiyoshi Sugiki and Kazuhiko Kato
University of Tsukuba

Abstract: We present a power-saving method for large-scale distributed storage systems. The key idea is to use virtual nodes and migrate them dynamically so as to skew the workload towards a small number of disks while not overloading them. Our proposed method consists of two kinds of algorithms, one for gathering or spreading virtual nodes according to the daily variation of workloads so that the active disks are reduced to a minimum, the other for coping with the changes in the popularity of data over a longer period. For this dynamic migration, data stored in virtual nodes are managed by a distributed hash table. Furthermore, to improve the reliability as well as to reduce the migration cost, we also propose an extension of our method by introducing a replication mechanism. The performance of our method is measured both by simulation and a prototype implementation. From the experiments, we observed that our method skews the workload so that the average load for the active physical nodes as a function of the overall capacity is 69%. At the same time, we maintain a preferred response time by setting a suitable maximum workload for each physical node.

Resource Allocation for Computing Independent Tasks in the Cloud with Budget Constraint

[slides]
Weiming Shi and Bo Hong
Georgia Institute of Technology

Abstract: We consider the problem of running a large amount of independent, equal-sized tasks in the cloud with budget constraint. We model the cloud infrastructure by a node-weighted edge-weighted star-shaped graph which captures the different computing power and communication capacity of the computing resources in the cloud. Instead of trying to minimize the makespan or the total completion time of the system, our study focuses on the maximization of the steady-state throughput of the system. We show that the specific budget-constrained steady-state throughput maximization problem can be formulated and solved as a linear programming problem.We identify two modes of the system, i.e., budget-bound and communication-bound where the closed-form solutions exist for the formulated problem.The best allocation scheme is benefit-first when the system is budget-bound: preference should be given to the nodes in the order of increasing cost; and is bandwidth-first when the system is communication-bound: preference should be given to compute nodes in the order of decreasing bandwidth.

Self-Organizing Agents for Service Composition in Cloud Computing

[slides]
J. Octavio Gutierrez-Garcia and Kwang-Mong Sim
Gwangju Institute of Science and Technology

Abstract: In Cloud service composition, collaboration between brokers and service providers is essential to promptly satisfy incoming Cloud consumer requirements. These requirements should be mapped to Cloud resources, whose access is provided by web services, in an automated manner. However, distributed and constantly changing Cloud-computing environments pose new issues to automated service composition such as: (i) dynamically contracting service providers, which set service fees on a supply-and-demand basis, and (ii) dealing with incomplete information regarding Cloud resources (e.g., location and providers). To address these issues, in this work an agent-based Cloud service composition approach is presented. Cloud participants and resources are implemented and instantiated by agents. These agents sustain a three-layered self-organizing multi-agent system that establishes a Cloud service composition framework and an experimental test bed. The self-organizing agents make use of acquaintance networks and the contract net protocol to evolve and adapt Cloud service compositions. The experimental results indicate that service composition is efficiently achieved despite dealing with incomplete information as well as coping with dynamic service fees.

HAMA: An Efficient Matrix Computation with the MapReduce Framework

[slides]
Sangwon Seo, Edward J. Yoon, Jaehong Kim, Seongwook Jin, Jin-Soo Kim, and Seungryoul Maeng
KAIST

Abstract: Various scientific computations have become so complex, and thus computation tools play an important role. In this paper, we explore the state-of-the-art framework providing high-level matrix computation primitives with MapReduce through the case study approach, and demonstrate these primitives with different computation engines to show the performance and scalability. We believe the opportunity for using MapReduce in scientific computation is even more promising than the success to date in the parallel systems literature.

The Two Quadrillionth Bit of Pi is 0! Distributed Computation of Pi with Apache Hadoop

[slides]
Tsz-Wo Sze
Yahoo! Inc.

Abstract: We present a new record on computing specific bits of pi, the mathematical constant, and discuss performing such computations on Apache Hadoop clusters. The specific bits represented in hexadecimal are

0E6C1294 AED40403 F56D2D76 4026265B CA98511D 0FCFFAA1 0F4D28B1 BB5392B8.

These 256 bits end at the 2,000,000,000,000,252nd bit position, which doubles the previous known record. The position of the first bit is 1,999,999,999,999,997 and the value of the two quadrillionth bit is 0.

The computation is carried out by a MapReduce program called DistBbp. To effectively utilize available cluster resources without monopolizing the whole cluster, we develop an elastic computation framework that automatically schedules computation slices, each a DistBbp job, as either map-side or reduce-side computation based on changing cluster load condition. We have calculated $\pi$ at varying bit positions and precisions, and one of the largest computations took 23 days of wall clock time and 503 years of CPU time on a 1000-node cluster.

Hybrid Map Task Scheduling for GPU-based Heterogeneous Clusters

[slides]
Koichi Shirahata, Hitoshi Sato, and Satoshi Matsuoka
Tokyo Institute of Technology

Abstract: RESEARCH MapReduce is a programming model that enables efficient massive data processing in large-scale computing environments such as supercomputers and clouds. Such large-scale computers tend to employ GPUs to enjoy its good peak performance and high memory bandwidth recently. Since it depends on running application characteristics and underlying computing environments, scheduling MapReduce tasks onto CPUs and GPUs for efficient execution is difficult. To address this problem, we have proposed a hybrid online scheduling technique for GPU-based computing clusters, which minimizes the execution time of a submitted job using dynamic profiles of Map tasks running on CPUs or GPUs. We implemented a prototype of our proposed scheduling technique by extending Hadoop. We did some experiments for this prototype using K-Means as a benchmark on a supercomputer. The results show that the proposed technique achieves 1.93 times faster than the Hadoop original scheduling algorithm at 64 nodes (1024 CPU cores and 128 GPU devices). The results also indicates that the performance of map tasks including both CPU tasks and GPU tasks is significantly affected by the overhead of map task invocation from Hadoop.

Pepper: An Elastic Web Server Farm for Cloud based on Hadoop

[slides]
Subramaniam Krishnan and Jean Christophe Counio
Yahoo! Inc.

Abstract: APPLICATION: Web application based processing is traditionally used to handle high throughput traffic. Web applications are hosted on server farms. However, providing application level scalability and isolation on such server farms has been a challenge. Using cloud serving infrastructures instead could potentially provide advantages such as scaling, centralized deployment and capacity planning. They also possess attractive qualities such as self-healing as well as ease in isolation and monitoring. The problem with applying this approach lies in the complicated nature and operational overhead of bootstrapping and operating cloud virtualization infrastructures. We present Pepper, a novel, simple, low cost and elastic web serving cloud platform built leveraging Hadoop and Zookeeper. The design of Pepper clearly demonstrates its ability to run in isolation different web applications and scale dynamically on a cluster of machines. Pepper is being successfully used in Yahoo! to run web applications that acquire and pre-process high frequency web feeds such as breaking news and finance quotes with minimal latency in order to retain content freshness with the ability to scale to millions of feeds every day.

Characterization of Hadoop Jobs using Unsupervised Learning

[slides]
Milind Bhandarkar, Shashank Phadke, and Sonali Aggarwal
Yahoo! Inc., Stanford University

Abstract: APPLICATION. MapReduce is a programming paradigm for parallel processing that is increasingly being used for data-intensive applications in cloud computing environments. An understanding of the characteristics of workloads running in MapReduce environments benefits both the service providers in the cloud and users. This work is based on characterizing Hadoop jobs running on Yahoo!'s production cluster using unsupervised learning. Unsupervised clustering techniques have been applied to many important problems. We use these techniques to cluster jobs that are similar in characteristics. Every hadoop job generates statistical counters like number of maps, reduces, file bytes read/written, HDFS bytes read/written etc. We use these counters and job configuration features like input format of the input/output file, type of compression used for the output file etc. to group the jobs. We study the centroid and density of these groupings. The centroid of these groups helps us in obtaining the characteristic job of each cluster. We also do a comparative analysis of the real production jobs and jobs simulated by our current benchmark tool - GridMix by comparing clusters of both. This is a useful study to establish a benchmark for performance of Hadoop workload.

SSS: An Implementation of Key-value Store based MapReduce Framework

[slides]
Hirotaka Ogawa, Hidemoto Nakada, Ryosei Takano, and Tomohiro Kudoh
AIST

Abstract: MapReduce has been very successful in implementing large-scale data-intensive applications. Because of its simple programming model, MapReduce has also begun being utilized as a programming tool for more general distributed and parallel applications, e.g., HPC applications. However, its applicability is limited due to relatively inefficient runtime performance and hence insufficient support for flexible workflow. In particular, the performance problem is not negligible in iterative MapReduce applications. On the other hand, today, HPC community is going to be able to utilize very fast and energy-efficient Solid State Drives (SSDs) with 10 Gbit/sec-class read/write performance. This fact leads us to the possibility to develop ``High-Performance MapReduce'', so called. From this perspective, we have been developing a new MapReduce framework called ``SSS'' based on distributed key-value store (KVS). In this paper, we first discuss the limitations of existing MapReduce implementations and present the design and implementation of SSS. Although our implementation of SSS is still in a prototype stage, we conduct two benchmarks for comparing the performance of SSS and Hadoop. The results indicate that SSS performs 1-10 times faster than Hadoop.

A Hybrid Distributed System Architecture for Storing and Processing Images from the Web

[slides]
Murali Krishna, Balaji Kannan, and Anand Ramani
Yahoo! Inc.

Abstract: Multimedia applications have undergone tremendous changes in the recent past that they have called for a scalable and reliable processing and storage framework. Image processing algorithms such as pornographic content detection becomes a lot more challenging in terms of accuracy, recall, and speed when run on billions of images. This paper presents the design and implementation of HMCS, a hybrid distributed architecture that uses Hadoop distributed file system for storage and Map/Reduce paradigm for processing images crawled from the web. This architecture combines the power of Hadoop framework when there is a need to parallelize the task as Map/Reduce jobs and uses stand alone crawler nodes to fetch relevant contents from the web. We also explain how the architecture scales to handle billions of images. Evaluations on real world web data indicate that the system can store and process billions of images in few hours.

Cogset vs. Hadoop: Measurements and Analysis

[slides]
Steffen Viken Valvåg, Åge Kvalnes, and Dag Johansen
University of Tromsø

Abstract: RESEARCH: Cogset is an efficient and generic engine for reliable storage and parallel processing of data. It supports a number of high-level programming interfaces, including a MapReduce interface compatible with Hadoop. In this paper, we evaluate Cogset’s performance as a MapReduce engine, comparing it to Hadoop. Our results show that Cogset generally outperforms Hadoop by a significant margin. We investigate the causes of this gap in performance and demonstrate some relatively minor modifications that markedly improve Hadoop’s performance, closing some of the gap.

Howdah - a flexible pipeline framework for analyzing genomic data

[slides]
Steven Lewis, Sheila Reynolds, Hector Rovera, Mike Oleary, Sarah Killcoyne, Ilya Shmulevich, and John Boyle
Institute For Systems Biology

Abstract: The advent of new high-throughput sequencing technologies has led to a flood of genomic data which overwhelms the capabilities of single processor machines. We present a MapReduce pipeline called Howdah that supports the analysis of genomic sequence data allowing multiple tests to be plugged in to a single MapReduce job. The pipeline is used to detect chromosomal abnormalities such as insertions, deletions and translocations as well as single nucleotide polymorphisms (SNPs).

Scaling Populations of a Genetic Algorithm for Job Shop Scheduling Problems using MapReduce

[slides]
Di-Wei Huang and Jimmy Lin
University of Maryland

Abstract: RESEARCH. Inspired by Darwinian evolution, a genetic algorithm (GA) approach is one of the popular heuristic methods for solving hard problems, such as the Job Shop Scheduling Problem (JSSP), which is one of the hardest problems where there lacks efficient exact solutions. It is intuitive that the population size of a GA may greatly affect the quality of the solution, but it is unclear what are the effects of having population sizes that are significantly greater than typical experiments. The emergence of MapReduce, a framework running on a cluster of computers that aims to provide large-scale data processing, offers great opportunities to investigate this issue. In this paper, a GA is implemented to scale the population using MapReduce. The experiments are conducted on a cluster of 414 machines, and population sizes up to 10^7 are inspected. It is shown that larger population sizes not only tend to find better solutions, but also require fewer generations. Therefore, it is clear that when dealing with a hard problem like JSSP, an existing GA can be improved by massively scaling up populations with MapReduce, such that it can be parallelized and completed in reasonable time.

A Study in Hadoop Streaming with Matlab for NMR data processing

[slides]
Kalpa Gunaratna, Paul Anderson , Ajith Ranabahu, and Amit P. Sheth
Knoesis Center, Air Force Research Lab

Abstract: Applying cloud computing techniques for analyzing large data sets has shown promise in many data-driven scientific applications. Our approach presented here is to use cloud computing for Nuclear Magnetic Resonance (NMR) data analysis which normally consists of large amount of data. Biologists often use third party or commercial software for ease of use. Enabling the capability to use this kind of a software in a cloud will be highly advantageous in many ways.

Scripting languages especially designed for clouds may not have the flexibility that biologists need for their purposes. But they are familiar with special software packages that allow them to write their complex calculations with minimum effort, which often are not compatible with a Cloud environment. Therefore, biologists who are trying to perform analysis on NMR data gets tremendous advantage of our proposed solution that gives them the flexibility to Cloud-enable their familiar software. This also enables them to perform calculations on huge amount of data that was not possible earlier. We are in the initial stage of developing a framework for NMR data analysis. Our study is also applicable to any other environment that needs similar flexibility.

A MapReduce-based architecture for rule matching in production system

[slides]
Bin Cao, Jianwei Yin, Qi Zhang, and Yanming Ye
Zhejiang University, Hangzhou, China

Abstract: Production system which accepts the facts and draws conclusions by repeatedly matching facts with rules plays an important role of improving the business by providing agility and flexibility. However, rule matching in production is badly time-consuming, and single computer limits the improvement for current matching algorithm. To address these problems, we proposed a MapReduce-based architecture to implement the distributed and parallel matching in different computers running with Rete algorithm. The architecture would benefit production system in efficient and performance, large scale of rules and facts are for special. This paper firstly formalizes some definitions for a accurate description, then not only discusses the details of implementation for different stages of the architecture but also shows the high efficiency through the experiment. At the end, we mention some complex factors which will be considered in the future for better performance.

Authorization as a Service for Cloud & SOA Applications

[slides]
Ulrich Lang
ObjectSecurity

Abstract: This paper introduces the concept of moving security and compliance policy automation for Cloud applications and mashups into the Cloud. The policy automation aspects covered in this paper include policy configuration, technical policy generation using model-driven security, application authorization management, and incident reporting. Policy configuration is provided as a subscription-based Cloud service to application develop-ment tools, and technical policy generation, enforcement and monitoring is embedded into Cloud application development and runtime platforms. OpenPMF Security & Compliance as a Service (“ScaaS”), a reference implemen-tation using ObjectSecurity OpenPMF, is also presented.

Security Services Lifecycle Management in On-Demand Infrastructure Services Provisioning

[slides]
Yuri Demchenko, Cees de Laat, Joan A. García-Espín and Diego Lopez
University of Amsterdam, I2CAT Foundation, RedIRIS

Abstract: Modern e-Science and high technology industry require high-performance and complicated network and computer infrastructure that should be provisioned on-demand to support distributed collaborating groups of researchers and applications. The effective use and management of the dynamically provisioned services can be achieved by using the Service Delivery Framework (SDF) proposed by TeleManagaement Forum that provides a good basis for defining the whole services life cycle management and supporting infrastructure services. The paper discusses conceptual issues, basic requirements and practical suggestions for provisioning consistent security services as a part of the general e-Science infrastructure provisioning, in particular Grid and Cloud based. The proposed Security Services Lifecycle Management (SSLM) model extends the existing frameworks with additional stages such as “Reservation Session B inding” and “Registration & Synchronisation” that specifically targets such security issues as the provisioned resources restoration or migration and provide a mechanism for remote data protection by binding them to the session context. The paper provides a short overview of the existing standards and technologies and refers to the on-going projects and experience in developing dynamic distributed security services.

Modeling the Runtime Integrity of Cloud Servers: a Scoped Invariant Perspective

[slides]
Jinpeng Wei, Calton Pu, Carlos V. Rozas and Anand Rajan
Florida International University, Georgia Institute of Technology, Intel Corporation

Abstract: One of the underpinnings of Cloud Computing security is the runtime integrity of individual Cloud servers. Due to the on-going discovery of runtime software vulnerabilities like buffer overflows, it is critical to be able to gauge the integrity of a Cloud server as it operates. In this paper, we propose scoped invariants as a primitive for analyzing the software system for its integrity properties. We report our experience with the modeling and detection of scoped invariants. The Xen Virtual Machine Manager is used for a case study. Our research detects a set of essential scoped invariants that are critical to the runtime integrity of Xen. One such property, that the addressable memory limit of a guest OS must not include Xen’s code and data, is indispensable for Xen’s guest isolation mechanism. The violation of this property demonstrates that the attacker only needs to modify a single byte in the Global Descriptor Table to achieve his goal.

Inadequacies of Current Risk Controls for the Cloud

[slides]
Sadie Creese, Mike Auty, Michael Goldsmith and Paul Hopkins
The University of Warwick

Abstract: In this paper we describe where current risk controls (as documented in ISO27001/27002) for mitigating information security risks are likely to be inadequate for use in the cloud. Such an analysis could provide a rationale for prioritizing protection research, and the work presented here is part of a larger exercise designed to identify the potential for cascade attacks in the cloud, and those areas most likely to be targeted based on both an understanding of threat motivations and likely areas of vulnerability.

A Privacy Impact Assessment Tool for Cloud Computing

[slides]
David Tancock, Siani Pearson and Andrew Charlesworth
HP Labs, University of Bristol

Abstract: In this paper, we present a Privacy Impact Assessment (PIA) decision support tool, that can be integrated - within a cloud environment. Privacy is an important issue in cloud computing, both in terms of legal compliance, security, and user trust. A PIA is a systematic process for evaluatingthe possible effects that a particular activity or proposal may have on an individuals privacy. It focuses on understanding the system, initiative or scheme, identifying and mitigating adverse privacy impacts, and informing decision makers in deciding whether the project should proceed and in what form. The PIA should be properly distinguished from other business processes such as privacy issue analysis, privacy audits, as these are applied to existing systems to ensure their continuing conformity with internal rules and external requirements.

A Framework for evaluating clustering algorithm

[slides]
Nehinbe Joshua
School of Computer Science & Electronic Engineering University of Essex, UK

Abstract: Clustering algorithms for reclassifying logs of intrusion detectors are not frequently evaluated despite the fact that the detectors isolate intrusions based on allowable and disallowable activities. The disallowable policy enforcers will alert only on events that are known to be bad while the allowable policy enforcer will alert on events that deviate from those that have been classified as good. However, these trade-offs become difficult to balance in a recent time due to the complexity of computer attacks. Accordingly, intrusion detectors generate tons of low level alerts that may signify realistic and false attacks. Importantly, clustering algorithms that reclassify audit logs erroneously process failed attacks. Also, the qualities of clustering schemes that the algorithms generate often lack objective evaluations. Consequently, attacks in progress are not forestalled despite the huge volume of low level warnings that intrusion detectors generate beforehand. Therefore, this paper presents category utility for evaluating clustering algorithm that was used to process audit trails. Series of evaluations showed how to adopt category utility to improve the efficacy of intrusion detections.

Do you get what you pay for? Using Proof-of-Work Functions to Verify Performance Assertions in the Cloud

[slides]
Falk Koeppe and Joerg Schneider
Technische Universitaet Berlin

Abstract: In the Cloud, the operators usually offer resources on a pay per use price model. The client gets access to a newly created virtual machine and has no direct access to the underlying hardware. Therefore, the client cannot verify whether the Cloud operator provides the negotiated amount of resources or only a fraction thereof. Especially, the assigned share of CPU time can be easily forged by the operator. The client could use a normal benchmark to verify the performance of his virtual machine. However, as the Cloud operator owns the underlying infrastructure, the operator could also tamper with the benchmark execution. We identified four attack vectors to modify the results of the benchmark. Based on these attack vectors, we showed that using proof-of-work functions can disable three of them. Proof-of-work functions are challenge response systems, where it is simple to generate a challenge and verify the result while solving the challenge is compute intensive. We implemented three proof-of-work functions in a prototype benchmark. Experiments showed that the runtime of the proof-of-work functions sufficiently relates to the results of the reference benchmark suite SPEC CPU2006.

Privacy, Security and Trust Issues Arising from Cloud Computing

[slides]
Siani Pearson and Azzedine Benameur
HP Labs

Abstract: Cloud computing is an emerging paradigm for large scale infrastructures. It has the advantage of reducing cost by sharing computing and storage resources, combined with an on-demand provisioning mechanism relying on a pay-per-use business model. These new features have a direct impact on the budgeting of IT budgeting but also affect traditional security, trust and privacy mechanisms. Many of the latter are no longer adequate, but need to be rethought to fit this new paradigm. In this paper we assess how security, trust and privacy issues occur in the context of cloud computing and propose mechanisms to address them.

CloudSEC: A Cloud Architecture for Composing Collaborative Security Services

[slides]
Jia Xu, Jia Yan, Liang He, Purui Su and Dengguo Feng
State Key Lab of Information Security, Institute of Software, CAS

Abstract: Massive Internet invasions implemented through the distributed platform fabricated by rapid diffusion of malwares, has become a significant issue in network security. We argue that the notion of “Collaborative Security” is an emerging trend in resisting distributed attacks originated from malware. Therefore, this paper proposes a new architecture: CloudSEC, for composing collaborative security-related services in clouds, such as correlated intrusion analysis, antispam, antiDDOS, automated malware detection and containment. CloudSEC is modeled as a dynamic peer-to-peer overlay hierarchy with three types of top-down architectural components. Based on this architecture, both data distribution and task scheduling overlays can be simultaneously implemented in a loosely coupled fashion, which can efficiently retrieve data resources from heterogeneous network security facilities, and harness distributed collection of computational resources to process data-intensive tasks. Hence, CloudSEC endues the network security infrastructure with the capability of dynamic adaptation and collaboration on an inter-organizational scale. The results of preliminary evaluation demonstrate that, CloudSEC not only deliver a sample service of distributed intrusion correlation with high scalability and robustness, but also achieves remarkable effectiveness in data sharing and task scheduling.

Trust and Cloud Services - An Interview Study

[slides]
Ilkka Uusitalo, Kaarina Karppinen, Juhola Arto and Reijo Savola
VTT Technical Research Centre of Finland

Abstract: The cloud services are on everybody’s lips, but what is the standpoint of those needing to consider the implications of serious involvement? What trust related aspects the experts deem noteworthy when surrendering organisation’s computing services into foreign hands? In this paper we present results from an expert interview study on trust and cloud services, covering the nature, measurement and experiencing of trust. Furthermore, future research directions regarding trust in the cloud are discussed based on the interviews.

Energy Use in the Media Cloud: Behaviour Change or Technofix?

[slides]
Chris Preist and Paul Shabajee
University of Bristol

Abstract: In this paper, we present an analysis of the potential worldwide demand for downloaded data and the resulting energy requirements of the cloud. Assuming that the average westerner’s media consumption moves fully online but does not rise substantially beyond current levels, and the global middle class reach western levels of consumption, we estimate the overall demand to be 3200MB/day per person, totaling 2570 Exabytes per year by the world population. We estimate the current energy demand for bandwidth to be 4Wh/MB, based on two independent sources of data. We conclude that the average power required to support this activity would be 1175GW at current levels of efficiency, and that a factor 60 performance improvement would be needed if infrastructure energy is to be provided by 1% of renewable energy capacity in 2030. By looking at historical trends in energy efficiency, we observe that this would be reached around 2021 if these trends continue. We document potential new applications that might require bandwidth capacity beyond our estimate. We also outline behavior change strategies that could be used to reduce the overall demand for bandwidth if historical performance improvements are not maintained, and as interim measures prior to efficiencies being realized.

Enabling Sustainable Clouds via Environmentally Opportunistic Computing

[slides]
Michał Witkowski, Paul R. Brenner, David B. Go, Ryan Jansen, and Eric M. Ward
Application Department, Poznan Supercomputing and Networking Center, University of Notre Dame

Abstract: Abstract Increasing economic and environmental costs of running data centers has driven interest in sustainable computing. Environmentally Opportunistic Computing (EOC) is an approach to harvesting heat produced by computing clusters. Our Green Cloud EOC prototype serves as an operational demonstration that IT resources can be integrated with the dominate energy footprint of existing facilities and dynamically controlled to balance process throughput, thermal transfer, and free cooling via job management and migration. We will describe the architecture and operation of this successful prototype with prime interest in the management of servers running in free cooling conditions.

Social Impact of Privacy in Cloud Computing

[slides]
Rui Maximo Esteves and Chunming Rong
University of Stavanger

Abstract: Cloud computing is emerging as a serious paradigm shift in the way we use computers. It relies over several technologies that are not new. However, the increasing availability of bandwidth allows new combinations and opens new IT perspectives. The data storage and processing power are being moved massively to more efficient and centralized structures over the web. The costs are being reduced with the loss of our data control as a trade-off. It will be almost inevitable for companies not following this trend. Yet, there are some important challenges to cross. This paper discusses Cloud Computing concept concerning privacy and how it may affect our freedom of speech.

On the Sustainability Impacts of Cloud-enabled Cyber Physical Space

[slides]
Tomasz Wiktor Wlodarczyk and Chunming Rong
University of Stavanger

Abstract: This paper establishes a relation between Cloud Computing and Cyber Physical Space. It demonstrates how Cloud Computing technologies provide necessary technical means to create Cyber Physical Space. At the same time it describes examples of cloud services that already create elements of the Cyber Physical Space. This relation is subsequently used as an important basis for analysis of sustainability impacts of Cloud Computing on human in society and nature. Due to significant growth of Cloud Computing a deep and critical discussion of those impacts is important. The main findings suggest that society is not equipped with right sociological and technical tools to handle the impacts of this rapid change, and that responsibility lays in the design that will ensure the sustainability through properly modeled interactions in Cyber Physical Space.

Framing the Issues of Cloud Computing & Sustainability: A Design Perspective

[slides]
Yue Pan, Siddharth Maini and Eli Blevis
Indiana University

Abstract: In this paper, we describe the present lack of understanding about if the potential environmental effects of transitions to cloud computing are positive or negative. We describe that research about the human interactivity implications of and for cloud computing has yet to enter the arena of Human Computer Interaction (HCI) in a significant way. We describe a short inventory of what is presently in the HCI literature apropos of cloud computing and interactivity. In addition, we offer a description of how we think the issues of cloud computing in the perspective of HCI may be framed, as well as an inventory of social issues implicated in cloud computing. Finally, we suggest some projects and problems that may be appropriate for advancing cloud computing in the perspective of HCI with sustainability as a key goal.

An Interface Design for Future Cloud-based Visualization Services

[slides]
Yuzuru Tanahashi, Cheng-Kai Chen, Stephane Marchesin, and Kwan-Liu Ma
University of California, Davis

Abstract: The pervasive concept of cloud computing suggests that visualization, which is both data and computing intensive, is a perfect cloud computing application. This paper presents a sketch of an interface design for an online visualization service. To make such a service attractive to a wider audience, its user interface must be simple and easy to use for both casual and expert users. We envision an interface that supports a visualization process mainly directed by browsing and assessing existing visualizations in terms of images and videos will be very appealing to, in particular, casual users. That is, the aim is to maximize the utilization of the rich visualization data on the web. Without losing generality, we consider volume data visualization applications for our interface design. We also discuss issues in organizing online visualization data, and constructing and managing a rendering cloud.

The Ethics of Cloud Computing: A Conceptual Review

[slides]
Bernd Carsten Stahl, Veikko Ikonen, and Job Timmermans, Engin Bozdag
De Montfort University, VTT, Technical University Delft

Abstract: Cloud computing can raise ethical issues. In many cases these will depend on particular applications and circumstances. The present paper sets out to identify ethical issues of cloud computing that arise from the fundamental nature of the technology rather than any specific circumstances. The paper describes how these general features were identified, how ethical issues arising from them were collected and it concludes by discussing means of addressing them.

User experience and Security in the Cloud – An Empirical Study in the Finnish Cloud Consortium

[slides]
Nilay Oza, Kaarina Karppinen, and Reijo Savola
VTT

Abstract: This paper presents an empirical analysis of user experience and security issues in cloud computing. The study is based on the rationale that superior user experience and user-centric security are the two crucial issues that help build a wholesome experience for the cloud service user. Qualitative research analysis is used to collect perspectives of eleven experts from Finnish Cloud Software Program Consortium. The perspectives of experts are then analyzed qualitatively and a thematic analysis is presented. As a result a range of issues related to user experience and security were identified, leading to a perspective of marrying security and user experience.

Cloud Computing for Enhanced Mobile Health Applications

[slides]
Fisseha Mekuria and Mzomuhle Nkosi
CSIR Council for Scientific & Industrial Research

Abstract: Mobile devices are being considered as service platforms for mobile health information delivery, access and communication. However mobiles face challenges with regard to delivering secure multimedia based health services due to limitations in computation and power supply. Since mobile devices have limited computational capacity and run on small batteries; they are unable to run heavy multimedia & security algorithms. In this paper a cloud computing framework to relieve mobile devices from executing heavier multimedia and security algorithms in delivering mobile health services is described. The proposed framework uses a Cloud Computing protocol management model which intends to provide multimedia sensor signal processing & security as a service to mobile devices. Our approach suggests that multimedia and security operations can be performed in the cloud, allowing mobile health service providers to subscribe and extend the capabilities of their mobile health applications beyond the existing mobile device limitations.

Personalized Modeling for SaaS Based on Extended WSCL

Ying Liu, Bin Zhang, Guoqi Liu, Deshuai Wang and Yichuan Zhang
Northeastern University, China

UTILIZING A SOCIAL NETWORKING SITE AS A WEB PORTAL TO PROCESS CRESIS RADAR DATA

Jeff Wood, Linda Hayden and Raminder Singh
Elizabeth City State University, Indiana University

ConTrail: Contents Delivery System Based on a Network Topology Aware Probabilistic Replication Mechanism

Yoshiaki Sakae, Masumi Ichien, Yasuo Itabashi, Takayuki Shizuno and Toshiya Okabe
System Platforms Research Laboratories, NEC Corporation, Japan

On-Demand Virtual Cluster in Cloud Web-Based OS Environment

Yi Lun Pan and ChangHsing Wu
National Center for High-performance Computing, Taiwan

Fault Tolerance for HPC with OpenVZ Virtualization by Lite Migration Toolkit

Yi Lun Pan and ChangHsing Wu
National Center for High-performance Computing, Taiwan

FU-JIN : A Cloud Computing Environment Visualization System for Specifying Points of Failure

Tomonori Ikuse, Shinya Kanda, Masato Jingu, Kyohei Moriyama, Masatoshi Enomoto, Hiroaki Hazeyama and Takeshi Okuda
Graduate School of Information Science, Nara Institute of Science and Technology, Japan

Horizontal Dynamic Cloud Collaboration Platform: Research Opportunities and Challenges

Mohammad Mehedi Hassan, Biao Song and Eui-Nam Huh
Kyung Hee University, Global Campus, Korea, Republic of

Standards Based Cloud Clients

Michael Behrens, Mark Carlson and David Moolenaar
R2AD, Oracle, R2AD, LLC

Elastic Scaling in the Clouds

Kate Keahey, John Bresnahan, Tim Freeman, David LaBissoniere and Paul Marshall
ANL, University of Chicago, University of Colorado, Boulder

HUBzero + Cloud: Power tool for the masses

Carol Song, David Braun, Christopher Thompson, Alex Younts and Preston Smith
Purdue University

AirNotes: a Location-aware Application in SaaS Pattern

Huiping Lin
Peking University, IBM CRL, China

SaaS Modeling Tool for Personalization

Ying Liu, bin zhang, guoqi liu, deshuai wang and yichuan zhang
Northeastern University, China

Dynamic Provisioned Experiments in FutureGrid

Gregor von Laszewski and Geoffrey Fox
Indiana University

Monitoring User-level Functionality and Performance using Inca

Shava Smallen
University of California, San Diego

Local Cloud Deployment in a Limited IP Setting

Michael Galloway
University of Alabama

Involuntary Computing: Hacking the Cloud

Sebastien Goasguen, Lance Stout and Michael Murphy
Clemson University, Coastal Carolina University

Efficient Bidding for Virtual Machine Instances in Clouds

Sharrukh Zaman and Daniel Grosu
Wayne State University

Performance Analysis of HPC Virtualization Technologies within FutureGrid

Andrew Younge, James Brown, Robert Henschel, Judy Qiu and Geoffrey Fox
Indiana University

An open source, cloud independent, Java API for high performance cloud computing application design, development, simulation and evaluation

Raffaele Montella, Francesca Lucarelli and Paolo Esposito
Department of Applied Science - University of Naples Parthenope, Italy

A Survery of Open-Source Cloud Infrastructure using FutureGrid Testbed

Tak-Lon Wu, Shakeela Noohir Baasha and Sonali Surendra Karwa
School of Informatics and Computing, Indiana University, Bloomington, IN

A Survey of Cloud Storage Systems

Xiaoming Gao, Pranav Shah, Adarsh Yoga, Abhijeet Kodgire and Xiaogang Ni
Indiana University

Improving Twister Messaging System Using Apache Avro

Yuduo Zhou, Patanachai Tangchaisin and Yuan Luo
Indiana University Bloomington

Memcached Integration with Twister

Saliya Ekanayake, Jerome Mitchell, Yiming Sun and Judy Qiu
Indiana University

The Study of Implementing PhyloD application with DryadLINQ

Adrija Sen, Ratul Bhawal and Chengming Ge
Indiana University

ModflowOnAzure: An On-demand “Groundwater Modeling as a Service” Solution

Yong Liu, Yan Xu and Wenming Ye
NCSA, Microsoft Research, Microsoft

Parallelism for LDA

Yang Ruan and Changsi An
Indiana University

Large Scale PageRank with MapReduce

Li Hui
Indiana University

On Using Cloud Platforms in a Software Architecture for Smart Energy Grids

Yogesh Simmhan
University of Southern California

Composable Services Architecture for Dynamically Configurable Virtualised Infrastructure Services Provisioning

Yuri Demchenko, Diego Lopez, Cees de Laat and Joan A. García-Espín
University of Amsterdam, Netherlands, RedIRIS, Spain, I2CAT Foundation, Spain

Survey on Cloud Computing Security

Shilpashree Srinivasamurthy and David Liu
Indiana University Purdue University Fort Wayne (IPFW)

TRIANGULATION THEORY: AN APPROACH TO MITIGATE GOVERNANCE RISKS IN CLOUDS

Rizwan Ahmad and Lech Janczewski
University of Auckland, New Zealand

Getting Code Near the Data: A Study of Generating Customized Data Intensive Scientific Workflows with Domain Specific Language

Ashwin Manjunatha, Ajith Ranabahu, Paul Anderson and Amit Sheth.
Kno.e.sis Center , Wright State University, Air Force Research Lab

Security Attacks and Solutions in Clouds

Kazi Zunnurhain and Susan Vrbsky
University of Alabama

MobiCloud - Making Clouds Reachable: A Toolkit for Easy and Efficient Development of Customized Cloud Mobile Hybrid Application

Ashwin Manjunatha, Ajith Ranabahu, Amit P. Sheth and Krishnaprasad Thirunarayan
Wright State University, Kno.e.sis Center , Wright State University

Map-Reduce Expansion of the ISGA Genomic Analysis Web Server

Chris Hemmerich, Adam Hughes, Yang Ruan, Aaron Buechlein, Judy Qiu and Geoffrey C. Fox
CGB and CGL, Indiana University

Parallel Applications And Tools For Cloud Computing Environments

Jong Youl Choi and Judy Qiu
Indiana University

Parallelizing a Computationally Intensive Financial R Application with Zircon Technology

Ron Guida
Zircon Computing, LLC

Elastic Application Platforms for Cloud Computing

Ron Guida and Doug Schmidt
Zircon Computing, LLC

Lyatiss Tuner

Pascale Vicat-blanc, sebastien soudan and guilherme Koslovski
Lyatiss, INRIA, France

Towards Secure Cloud Storage

SeongHan Shin and Kazukuni Kobara
Northeastern University, China

Facilitating access to customized computational infrastructure for plant sciences: Atmosphere cloudfront

Seung-jin Kim, Nirav Merchant and Edwin Skidmore
AIST, Japan








Organized by Indiana University
Sponsored by the IEEE Computer Society - Technical Committee of Scalable Computing (TCSC)
ClouldCom2010 Sponsors