Energy optimization techniques for virtualized clouds: Enhanced power modeling & energy-aware VM placements

In recent years, the datacenters have observed unprecedented growth both in terms of size as well as magnitude. Virtualization solutions are widely implemented to solve various problems of modern day datacenters which include: hardware underutilization, data center space usage and high system administration and maintenance cost. The major challenges faced by these large server farms are the lack of high system reliability and the high operational cost due to large electricity consumption. Thus, energy-aware VM deployments and scheduling is an urgent necessity to achieve both of these goals. Job scheduling has been studied by researches for several years now, but the development of virtualized clusters and cloud environments has opened door for new approaches to job scheduling. The major components of job scheduling in virtualized environment consist of: VM placement among the available physical resources and the dynamic workload balancing with the help of job migrations across datacenter cluster nodes.;There are several ongoing research efforts towards the development of an integrated cloud management system to provide comprehensive online monitoring of resources utilization along with the implementation of power-aware policies to reduce the total energy consumption. However, most of these techniques provide online power monitoring based on the power consumption of a physical node running one or more Virtual Machines (VM). They lack a fine-grained mechanism to profile the power of an individual hosted VM. In this work we propose a novel power modelling technique, VMeter, based on online monitoring of system-resources having high correlation with the total power consumption. The monitored system sub-components include: CPU, cache, disk, and DRAM. The proposed model predicts instantaneous power consumption of an individual VM hosted on a physical node besides the full system power consumption. Our model is validated using computationally diverse and industry standard benchmark programs. Our evaluation results show that our model is able to predict instantaneous power with an average mean and median accuracy of 93% and 94%, respectively, against the actual measured power using an externally attached power meter. With a fine grained monitoring infrastructure in place, we then propose two energy-aware VM placement policies namely: Performance Aware and Load-Factor Aware deployments. Both these policies analyze the power state information of each physical node within the cloud to determine the optimal set of physical nodes to be used to deploy new VMs. The simulation results show that both of our proposed algorithms lead to substantial reduction in the total energy consumption compared to the first-fit VM placement strategy. The load-factor aware algorithm performs better than the performance-aware algorithm for most of the test scenarios. Moreover, our simulation results provide a comprehensive analysis of the datacenter total energy consumption by varying job parameters including: job length (large, medium and small jobs), job nodes requirement, and job priority etc.