Abstract
Cloud computing provides on-demand access to a shared pool of computing resources, which enables organizations to outsource their IT infrastructure. Cloud providers are building data centers to handle the continuous increase in cloud users' demands. Consequently, these cloud data centers consume, and have the potential to waste, substantial amounts of energy. This energy consumption increases the operational cost and the CO2 emissions. The goal of this paper is to develop an optimized energy and SLA-aware virtual machine (VM) placement strategy that dynamically assigns VMs to Physical Machines (PMs) in cloud data centers. This placement strategy co-optimizes energy consumption and service level agreement (SLA) violations. The proposed solution adopts utility functions to formulate the VM placement problem. A genetic algorithm searches the possible VMs-to-PMs assignments with a view to finding an assignment that maximizes utility. Simulation results using CloudSim show that the proposed utility-based approach reduced the average energy consumption by approximately 6 % and the overall SLA violations by more than 38 %, using fewer VM migrations and PM shutdowns, compared to a well-known heuristics-based approach.
Highlights
Cloud computing delivers application, platform or infrastructure services to large numbers of users with diverse and dynamically changing requirements
In “Optimizing virtual machine placement” section, we provide a precise description of the VM placement problem
Comparing the results from Configuration 2.1 and Configuration 2.2, we can conclude that increasing the number of VMs by 50 % using the utility based approach resulted in nearly the same percentage of overall service level agreement (SLA) violations, while the energy consumption is increased by about 50 %
Summary
Platform or infrastructure services to large numbers of users with diverse and dynamically changing requirements. We focus on the problem of adaptively allocating virtual machines (VMs) to physical hosts, in the context of unpredictable workloads This involves making decisions such as when to relocate VMs, which VMs to relocate, where to place VMs that are to be relocated, and which physical machines can be switched off. In a series of papers Beloglazov et al [2,3,4] develop heuristic algorithms that make dynamic workload allocation decisions, taking into account energy usage when deciding where to place VMs. In adopting a heuristic approach, these papers focus on identifying criteria that suggest that an adaptation may be beneficial (which tends to involve detecting which hosts are over- or underloaded), and making reallocation decisions in ways that take into account estimated energy usage. By systematically refining their heuristics, the authors were able to make proposals that significantly improved on their static counterparts
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
