Jean-Marc Menaud

Entropy: a consolidation manager for clusters

Clusters provide powerful computing environments, but in practice much of this power goes to waste, due to the static allocation of tasks to nodes, regardless of their changing computational requirements. Dynamic consolidation is an approach that migrates tasks within a cluster as their computational requirements change, both to reduce the number of nodes that need to be active and to eliminate temporary overload situations. Previous dynamic consolidation strategies have relied on task placement heuristics that use only local optimization and typically do not take migration overhead into account. However, heuristics based on only local optimization may miss the globally optimal solution, resulting in unnecessary resource usage, and the overhead for migration may nullify the benefits of consolidation. In this paper, we propose the Entropy resource manager for homogeneous clusters, which performs dynamic consolidation based on constraint programming and takes migration overhead into account. The use of constraint programming allows Entropy to find mappings of tasks to nodes that are better than those found by heuristics based on local optimizations, and that are frequently globally optimal in the number of nodes. Because migration overhead is taken into account, Entropy chooses migrations that can be implemented efficiently, incurring a low performance overhead.

Autonomic virtual resource management for service hosting platforms

Cloud platforms host several independent applications on a shared resource pool with the ability to allocate computing power to applications on a per-demand basis. The use of server virtualization techniques for such platforms provide great flexibility with the ability to consolidate several virtual machines on the same physical server, to resize a virtual machine capacity and to migrate virtual machine across physical servers. A key challenge for cloud providers is to automate the management of virtual servers while taking into account both high-level QoS requirements of hosted applications and resource management costs. This paper proposes an autonomic resource manager to control the virtualized environment which decouples the provisioning of resources from the dynamic placement of virtual machines. This manager aims to optimize a global utility function which integrates both the degree of SLA fulfillment and the operating costs. We resort to a Constraint Programming approach to formulate and solve the optimization problem. Results obtained through simulations validate our approach.

SLA-Aware Virtual Resource Management for Cloud Infrastructures

Cloud platforms host several independent applications on a shared resource pool with the ability to allocate computing power to applications on a per-demand basis. The use of server virtualization techniques for such platforms provide great flexibility with the ability to consolidate several virtual machines on the same physical server, to resize a virtual machine capacity and to migrate virtual machine across physical servers. A key challenge for cloud providers is to automate the management of virtual servers while taking into account both high-level QoS requirements of hosted applications and resource management costs. This paper proposes an autonomic resource manager to control the virtualized environment which decouples the provisioning of resources from the dynamic placement of virtual machines. This manager aims to optimize a global utility function which integrates both the degree of SLA fulfillment and the operating costs. We resort to a Constraint Programming approach to formulate and solve the optimization problem. Results obtained through simulations validate our approach.

Performance and Power Management for Cloud Infrastructures

A key issue for Cloud Computing data-centers is to maximize their profits by minimizing power consumption and SLA violations of hosted applications. In this paper, we propose a resource management framework combining a utility-based dynamic Virtual Machine provisioning manager and a dynamic VM placement manager. Both problems are modeled as constraint satisfaction problems. The VM provisioning process aims at maximizing a global utility capturing both the performance of the hosted applications with regard to their SLAs and the energy-related operational cost of the cloud computing infrastructure. We show several experiments how our system can be controlled through high level handles to make different trade-off between application performance and energy consumption or to arbitrate resource allocations in case of contention.

An expressive aspect language for system applications with Arachne

C applications, in particular those using operating system level services, frequently comprise multiple crosscutting concerns: network protocols and security are typical examples of such concerns. While these concerns can partially be addressed during design and implementation of an application, they frequently become an issue at runtime, e.g., to avoid server downtime. A deployed network protocol might not be efficient enough and may thus need to be replaced. Buffer overflows might be discovered that imply critical breaches in the security model of an application. A prefetching strategy may be required to enhance performance.While aspect-oriented programming seems attractive in this context, none of the current aspect systems is expressive and efficient enough to address such concerns. This paper presents a new aspect system to provide a solution to this problem. While efficiency considerations have played an important part in the design of the aspect language, the language allows aspects to be expressed more concisely than previous approaches. In particular, it allows aspect programmers to quantify over sequences of execution points as well as over accesses through variable aliases. We show how the former can be used to modularize the replacement of network protocols and the latter to prevent buffer overflows. We also present an implementation of the language as an extension of Arachne, a dynamic weaver for C applications. Finally, we present performance evaluations supporting that Arachne is fast enough to extend high performance applications, such as the Squid web cache.

Power management in grid computing with xen

While chip vendors still stick to Moores law, and the performance per dollar keeps going up, the performance per watt has been stagnant for last few years. Moreover energy prices continue to rise worldwide. This poses a major challenge to organisations running grids, indeed such architectures require large cooling systems. Indeed the one-year cost of a cooling system and of the power consumption may outfit the grid initial investment. We observe, however, that a grid does not constantly run at peak performance. In this paper, we propose a workload concentration strategy to reduce grid power consumption. Using the Xen virtual machine migration technology, our power management policy can dispatch transparently and dynamically any applications of the grid. Our policy concentrates the workload to shutdown nodes that are unused with a negligible impact on performance. We show through evaluations that this policy decreases the overall power consumption of the grid significantly.

Web cache prefetching as an aspect: towards a dynamic-weaving based solution

Given the high proportion of HTTP traffic in the Internet, Web caches are crucial to reduce user access time, network latency, and bandwidth consumption. Prefetching in a Web cache can further enhance these benefits. For the best performance, however, the prefetching policy must match user and Web application characteristics. Thus, new prefetching policies must be loaded dynamically as needs change.Most Web caches are large C programs, and thus adding one or more prefetching policies to an existing Web cache is a daunting task. The main problem is that prefetching concerns crosscut the cache structure. Aspect-oriented programming is a natural technique to address this issue. Nevertheless, existing approaches either do not provide dynamic weaving, incur a high overhead for invocation of dynamically loaded code, or do not target C applications. In this paper we present μ-Dyner, which addresses these issues. In particular, μ-Dyner provides a low overhead for aspect invocation, that meets the performance needs of Web caches.

Cluster-wide context switch of virtualized jobs

Clusters are mostly used through Resources Management Systems (RMS) with a static allocation of resources for a bounded amount of time. Those approaches are known to be insufficient for an efficient use of clusters. To provide a finer RMS, job preemption, migration and dynamic allocation of resources are required. However due to the complexity of developing and using such mechanisms, advanced scheduling strategies have rarely been deployed. This trend is currently evolving thanks to the use of migration and preemption capabilities of Virtual Machines (VMs). However, although the manipulation of jobs composed of VM enables to change the state of the jobs according to the scheduling objective, changing the state and the location of numerous VMs at each decision is tedious and degrades the overall performance. In addition to the scheduling policy implementation, developers have to focus on the feasibility of the actions while executing them in the most efficient way. In this paper, we argue such an operation is independent from the policy itself and can be addressed through a generic mechanism, the cluster-wide context switch. Thanks to it, developers can implement sophisticated algorithms to schedule jobs without handling the issues related to their manipulations. They only focus on the implementation of their algorithm to select the jobs to run while the cluster-wide context switch system performs the necessary actions to switch from the current to the new situation. As a proof of concept, we evaluate the interest of the cluster-wide context switch through a sample scheduler that executes jobs as early as possible, even partially, regarding to their current resources requirements and their priority.

A New Protocol for Efficient Cooperative Transversal Web Caching

The bandwidth demands on the World Wide Web continue to grow at an exponential rate. To address this problem, many research activities are focusing on the design of Web caches. Unfortunately, Web caches exhibit poor performance with a hit rate of about 30%. A solution to improve this rate, consists of groups of cooperating caches. In its most general form, a cooperative cache system includes protocols for hierarchical and transversal caching. Although such a system brings better performance, its drawback lies in the resulting network load due to the number of messages that need to be exchanged to locate an object. This paper introduces a new protocol for transversal cooperative caching, which significantly reduces the associated network load compared to that induced by existing systems. ...

Constructing component-based extension interfaces in legacy systems code

Implementing an extension of a legacy operating system requires knowing what functionalities the extension should provide and how the extension should be integrated with the legacy code. To resolve the first problem, we propose that the use of a component model can make explicit the interface between an extension and legacy code. To resolve the second problem, we propose to augment interface specifications with rewrite rules that integrate support for extensions in the legacy code. We illustrate our approach using extensions that add new scheduling policies to Linux and prefetching to the Squid Web cache. In both cases a small number of rules are sufficient to describe modifications that apply across the implementation of a large legacy system.