Modeling and Analysis of Queueing-Based Vary-OnVary-Off Schemes for Server Clusters

Abstract EN

A cloud system usually consists of a lot of server clusters handling various applications.

Tosatisfy the increasing demands, especially for the front-end web applications, the computingcapacity of a cloud system is often allocated for the peak demand.

Such installation causesresource underutilization during the off-peak hours.

Vary-On/Vary-Off (VOVO) schemes concentrateworkloads on some servers instead of distributing them across all servers in a clusterto reduce idle energy waste.

Recent VOVO schemes adopt queueing theory to model the arrivalprocess and the service process for determining the number of powered-on servers.

For the arrivalprocess, Poisson process can be safely assumed in web services due to the large numberof independent sources.

On the other hand, the heavy-tailed distribution of service times is observedin real web systems.

However, there are no exact solutions to determine the performancefor M/heavy-tailed/m queues.

Therefore, this paper presents two queueing-based sizing approximationsfor Poisson and non-Poisson governed service processes.

The simulation resultsof the proposed approximations are analyzed and evaluated by comparing with the simulatedsystem running at full capacity.

This relative measurement indicates that the Pareto distributedservice process may be adequately modeled by memoryless queues when VOVO schemes areadopted.