Monte Carlo Based Server Consolidation for Energy Efficient Cloud Data Centers

Abstract:

The growing energy consumption of data centers is a compelling global problem and effective server consolidation is at the heart of energy efficient cloud data centers. Through efficient allocation of computing resources, we can consolidate virtual resources into as few physical servers as possible, with the aim of making the data center's power usage proportional to demand.

A variant of bin packing can be used to model the server consolidation problem, where the constraints are multidimensional and heterogeneous vectors rather than scalars and the goal is to satisfy the requested resource allocation using the minimum number physical servers. Since bin packing is NP-hard, we must rely on heuristics for practical solutions.  Variations of First Fit Decreasing (FFD) based heuristics have been shown to be effective both in theory and in practice for the single dimensional homogeneous case. However, the multidimensional and heterogeneous aspects of the server consolidation problem make it more complicated, requiring additional research to adapt FFD to the server consolidation problem.  We present existing server consolidation heuristics and a new FFD-based technique based on Monte Carlo simulation and Shannon entropy, which considers resource bottlenecks and dynamically adjusts to utilization variance of different resources.

Biography:

Bryan Harris is currently a PhD student at the Department of Computer Eng. & Computer Science at the University of Louisville, where he previously received his Master of Engineering.  He works with Dr. Nihat Altiparmak in the Computer Systems Lab on high performance storage and distributed computing topics.  Current research investigates the system-level impacts of new generation storage technologies, such as ultra low latency storage devices (e.g. Intel's 3D XPoint, Samsung's Z-NAND, etc.) and host controlled ("open channel") SSDs; work towards self-optimizing storage systems; improvements to parallel storage for high performance computing using Lustre; and studies into performance and energy efficiency improvements in distributed computing environments.