Abstract not found

Abstract. Cloud computing is receiving increasingly attention as it provides infinite resource capacity and “pay-as-you-go ” resource usage pattern to hosted applications. To maintain its SLA targets, resource provisioning of service-oriented applications in the cloud requires reliable performance from the cloud resources. In this paper, we study performance behavior of small instances in Amazon EC2. We demonstrate that the performance of virtual instances is relatively stable over time with fluctuations of mean response time within at most 8 % of the long-term average. Moreover, we also show that different supposedly identical instances often have very different performance, up to a ratio 4 from each other. We consider this as an important issue that must be addressed, but also as an opportunity as it allows one to assign each instance with a task that matches its own performance profile. 1

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. I.

Abstract—Today’s highly successful Massively Multiplayer Online Games (MMOGs) have millions of registered users and hundreds of thousands of active concurrent users. As a result of the highly dynamic MMOG usage patterns, the MMOG operators pre-provision and then maintain throughout the lifetime of the game tens of thousands of compute resources in data centers located across the world. Until recently, the difficulty of porting the MMOG software services to different platforms made it impractical to dynamically provision resources external to the MMOG operators ’ data centers. However, virtualization is a new technology that promises to alleviate this problem by providing a uniform computing platform with minimal overhead. To investigate the potential of this new technology, in this paper we propose a new hybrid resource provisioning model that uses a smaller and less expensive set of self-owned data centers, complemented by virtualized cloud computing resources during peak hours. Using real traces from RuneScape, one of the most successful contemporary MMOGs, we evaluate with simulations the effectiveness of the on-demand cloud resource provisioning strategy for MMOGs. We assess the impact of provisioning of virtualized cloud resources, analyze the components of virtualization overhead, and compare provisioning of virtualized resources with direct provisioning of data center resources. I.

Abstract—Many scientists perform extensive computations by executing large bags of similar tasks (BoTs) in mixtures of computational environments, such as grids and clouds. Although the reliability and cost may vary considerably across these environments, no tool exists to assist scientists in the selection of environments that can both fulfill deadlines and fit budgets. To address this situation, we introduce the ExPERT BoT scheduling framework. Our framework systematically selects from a large search space the Pareto-efficient scheduling strategies, that is, the strategies that deliver the best results for both makespan and cost. ExPERT chooses from them the best strategy according to a general, user-specified utility function. Through simulations and experiments in real production environments, we demonstrate that ExPERT can substantially reduce both makespan and cost in comparison to common scheduling strategies. For bioinformatics BoTs executed in a real mixed grid+cloud environment, we show how the scheduling strategy selected by ExPERT reduces both makespan and cost by 30%-70%, in comparison to commonlyused scheduling strategies. Keywords—bags-of-tasks; cloud; grid; Pareto-frontier I.

Abstract—Many applications in federated Grids have qualityof-service (QoS) constraints such as deadline. Admission control mechanisms assure QoS constraints of the applications by limiting the number of user requests accepted by a resource provider. However, in order to maximize their profit, resource owners are interested in accepting as many requests as possible. In these circumstances, the question that arises is: what is the effective number of requests that can be accepted by a resource provider in a way that the number of accepted external requests is maximized and, at the same time, QoS violations are minimized. In this paper, we answer this question in the context of a virtualized federated Grid environment, where each Grid serves requests from external users along with its local users and requests of local users have preemptive priority over external requests. We apply analytical queuing model to address this question. Additionally, we derive a preemptionaware admission control policy based on the proposed model. Simulation results under realistic working conditions indicate that the proposed policy improves the number of completed external requests (up to 25%). In terms of QoS violations, the 95 % confidence interval of the average difference with other policies is between (14.79%,18.56%). (external users) through contracts between them [4]. On the other hand, local users of each resource provider (cluster) send their requests directly to the local resource manager (LRM) of the cluster. Typically, local requests have priority over external requests in each cluster [7]. In other words, the organization that owns the resources would like to ensure that its local users (hereafter termed local requests) have priority access to the resources. In this situation, external users (hereafter termed external requests) are welcome to use resources if they are available. Nonetheless, external requests should not delay the execution of local requests. I.

With the increasing popularity of Cloud computing, the requirement for services supporting brokering across multiple infrastructure providers is growing rapidly. Cloud Computing environments are not only dynamic, but also heterogeneous with multiple types of Virtual Machine (VM) offered by various infrastructure providers. Similarly, the demand on services can also vary with time, which affects the number of VMs to be initiated. In this environment, the aim of Software as a Service (SaaS) providers is to maximize their profit and enhance their reputation by meeting Service Level Agreement (SLA) requirements of all accepted requests. SLAs are signed between SaaS providers and the customers to decide on the issues such as payment and Quality of Service (QoS). Thus, SaaS providers need effective strategies for accepting particular request, how many and what type of VMs to be initiated from suitable IaaS provider. This paper proposes admission control and scheduling algorithms that take into account dynamic parameters such as variation in VM’s initiation time and user’s QoS requirements such as budget deadline, and penalty rate ratio. This paper also presents an extensive evaluation study to analyse well suited algorithm for a particular scenario to maximise the SaaS provider's profit. Keywords: Cloud computing; Service Level Agreement (SLA); Admission Control; Software as a Service;

(MMOGs) have millions of registered users and hundreds of thousands of active concurrent players. Motivated by performance and portability, MMOG operators currently pre-provision and then maintain throughout the lifetime of the game tens of thousands of compute resources in data centers located across the world. In contrast, virtualization is a new technology that promises minimal overhead and a uniform, easy to port to, computing platform. To investigate the potential of this new technology, in this paper we propose a new hybrid resource provisioning model that uses a smaller and less expensive set of self-owned data centers, complemented by virtualized cloud computing resources during peak hours. Through trace-based simulation and empirical experimentation, we assess the impact of provisioning virtualized cloud resources, analyze the components of virtualization overhead, and compare provisioning of virtualized resources with direct provisioning of data center resources. Using a simple cost model, we also investigate the costs of hosting MMOGs on the resources leased independently from three commercial cloud providers, including the current market leader Amazon. We find that the virtualization characteristics, policy, and provider can have an important impact on the performance and cost of MMOG operation. We also find that resource under-allocation, which occurs more often for static resource provisioning and non-virtualized platforms, has a significant impact on a key performance metric, latency. Copyright c ○ 2009 Inderscience Enterprises Ltd. 2 A. Iosup, V. Nae, and R. Prodan

Overlay networking with a layer 2 abstraction provides a powerful model for virtualized wide-area distributed computing resources, including for high performance computing (HPC) on collections of virtual machines (VMs). With the emergence of cloud computing, it is now possible to allow VMs hosting tightly-coupled HPC applications to seamlessly migrate between distributed cloud resources and tightly-coupled supercomputing and cluster resources. However, to achieve the application performance that the tightly-coupled resources are capable of, it is important that the overlay network not introduce significant overhead relative to the native hardware. To support such a model, we describe the design, implementation, and evaluation of a virtual networking system that has negligible latency and bandwidth overheads in 1-10 Gbps networks. Our system, VNET/P, is embedded into our publicly available Palacios virtual machine monitor (VMM). VNET/P achieves native performance on 1 Gbps Ethernet networks and very high performance on 10 Gbps Ethernet networks and InfiniBand. On the latter networks, performance is converging to native speeds as we continue to enhance VNET/P. In comparison to VNET/U, its previously demonstrated, and highly optimized, user-level counterpart, VNET/P can achieve bandwidths over 33 times as high. These results suggest that it is feasible to extend a software-based overlay network designed to facilitate computing at wide-area scales into tightly-coupled environments.

Abstract — With the advent of data-intensive applications that generate large volumes of real-time data, distributed stream processing systems (DSPS) become increasingly important in domains such as social networking and web analytics. In practice, DSPSs must handle highly variable workloads caused by unpredictable changes in stream rates. Cloud computing offers an elastic infrastructure that DSPSs can use to obtain resources on-demand, but an open problem is to decide on the correct resource allocation when deploying DSPSs in the cloud. This paper proposes an adaptive approach for provisioning virtual machines (VMs) for the use of a DSPS in the cloud. We initially perform a set of benchmarks across performance metrics such as network latency and jitter to explore the feasibility of cloud-based DSPS deployments. Based on these results, we propose an algorithm for VM provisioning for DSPSs that reacts to changes in the stream workload. Through a prototype implementation on Amazon EC2, we show that our approach can achieve low-latency stream processing when VMs are not overloaded, while adjusting resources dynamically with workload changes. I.