ISim: A Novel Power Aware Discrete Event Simulation Framework for Dynamic Workload Consolidation and Scheduling in Infrastructure Clouds

Today’s cloud environment is hosted in mega datacenters and many companies host their private cloud in enterprise datacenters. One of the key challenges for cloud computing datacenters is to maximize the utility of the Processing Elements (PEs) and minimize the power consumption of the applications hosted on them. In this paper we propose a framework called ISim, wherein a Datacenter manager playing the role of a Meta-scheduler minimizes power consumption by exploiting different power saving states of the processing elements. The considered power management techniques by the ISim framework are dynamic workload consolidation and usage of low power states on the processing elements. The meta-scheduler aims at maximizing the utility of the cores by performing dynamic workload consolidation using context switching between the cores inside the chip. The Datacenter manager makes use of a prediction algorithm to predict the number of cores that are required to be kept in active state to fulfil the input service request at a given moment, thus maximizing the CPU utilization. The simulation results show, how power can be conserved from the host level till the core level in a datacenter with the optimal usage of different power saving states without compromising the performance.