PROJECT TITLE :

Data Center Server Provision: Distributed Asynchronous Control for Coupled Renewal Systems - 2017

ABSTRACT:

This paper considers a price minimization downside for knowledge centers with N servers and randomly arriving service requests. A central router decides that server to use for each new request. Each server has 3 varieties of states (active, idle, and setup) with totally different prices and time durations. The servers operate asynchronously over their own states and will opt for one of multiple sleep modes when idle. We have a tendency to develop an on-line distributed management algorithm thus that every server makes its own decisions. The request queues are bounded and the overall time average value is near optimal with probability one. First the algorithm will not would like likelihood info for the arrival rate or job sizes. Finally, an improved algorithm that uses one queue is developed via a “virtualization” technique, which is shown to supply the identical (near optimal) prices. Simulation experiments on a true information center traffic trace demonstrate the efficiency of our algorithm compared with alternative existing algorithms.