A Holistic Model of the Performance and the Energy-Efficiency of Hypervisors in an HPC Environment

Abstract

SUMMARYVirtualization is emerging as the prominent approach to mutualise the energy consumed by a single server running multiple Virtual Machines (VMs) instances. The efficient utilization of virtualized servers and/or computing resources requires understanding of the overheads in energy consumption and the throughput, especially on high-demanding High Performance Computing (HPC) platforms. In this paper, a novel holistic model for the power of virtualized computing nodes is proposed. Moreover, we create and validate instances of the proposed model using concrete measures taken during a benchmarking process that reflects an HPC usage, i.e. HPCC, IOZone and Bonnie++, conducted using two different hardware configurations on Grid5000 platform, based on Intel and AMD processors, and three widespread virtualization frameworks, namely Xen, KVM, and VMware ESXi. The proposed holistic model of machine power takes into account the impact of utilisation metrics of the machine's components, as well as the employed application, virtualization, and hardware. The model is further derived using tools such as multiple linear regressions or neural networks that prove its elasticity, applicability and accuracy. The purpose of the model is to enable the estimation of energy consumption of virtualized platforms, aiming to make possible the optimization, scheduling or accounting in such systems, or their simulation.

A Holistic Model for Resource Representation in Virtualized Cloud Computing Data Centers

“…Simulation of the resource usage is important not only for assuring the correct execution of applications and platform management, but also for proper power accounting. Our previous experimental study [2] confirms an increased accuracy of power estimation while using larger set of predictors [3]. The applicability of the model is tested by assessing its overhead as well as presenting the expressiveness of its features.…”

“…The proposed model is based on the observations that the power consumed by the system, as well as its performance, depends on the load levels of the hardware components. The first introduction of the proposed holistic model was focused on individual servers and experimental evaluation in terms of system power consumption and performance [2].…”

“…This flexibility can be also used to simulate the influence of the prospective hardware changes on the data center operation. The detailed information about utilization of resources can be used to feed more accurate power models, as increasing the set of predictors leads to a more accurate estimation of power [2].…”

Comparisons of Heat Map and IFL Technique to Evaluate the Performance of Commercially Available Cloud Providers

“…are considered in CloudCmp [11] to compare the performance of different Cloud services such as Amazon EC2, Windows Azure and Rackspace. Such low-level performance metrics could be further used to create models of high-level system properties, such as power consumption or performance [12], but it has not been included in CloudCmp. Qu et al [13] have proposed a cloud service selection process based on consumer experience and involving the third party to avoid a biased assessment of cloud services from users, however, it does not cover the performance measurements from cloud providers.…”

Using data-flow analysis in MAS for power-aware HPC runs

“…More precisely, the DFG representation allows each agent to evaluate and anticipate the incoming tasks to be executed. Coupled with a holistic modelling of the HPC platform introduced in a previous article [5] and adapted to the latest hardware advances, work-stealing operations and local decisions as regards the power state of the computing resources can be taken by the agents in order to dynamically optimize the power-efficiency of the execution, if possible without degrading the computing performance.…”