A Self-Optimizing Workload Management Solution for Cloud Applications

Given the dynamic nature of the Cloud, resulting from mapping virtual to physical resources, changes in the usage pattern of resources, and migration of virtual resources, in addition to the dynamic nature of the applications themselves, the bottleneck resource in a given application changes over time. Whether the bottleneck resource is a hardware or software resource, physical or virtual, the performance of the application deteriorates accordingly, leading to longer response times and saturated throughput. Today, a typical design of a workload manager for an application manages to a target performance measure, e.g. response time, and/or a system related performance measure, e.g. CPU utilization. We claim that, due to the dynamics of the Cloud and the unavailability of credible resource performance measures, an application workload manager needs to perform its objective in the absence of such target values. Moreover, the application workload manager should be able to function with a model-free controller, thus avoiding the complexity of dynamically changing its model as the cloud environment, especially the bottleneck resource, where the application is hosted and managed changes. In this paper we focus on an application workload manager which uses admission control as a means for load management. We present a design for such a self-optimizing workload manager that is both target-less and model-free. Our approach is to devise black-box bottleneck analytic techniques, combined with a simple binary controller. We demonstrate the validity of our workload manager in an experimental setup, using the RUBiS web application benchmark.

By: Haishan Wu, Asser N. Tantawi, Tao Yu

Published in: Proceedings of 2013 IEEE 20th International Conference on Web Services (ICWS)Los Alamitos, CA,IEEE Computer Society,, p.483-90 in 2013

LIMITED DISTRIBUTION NOTICE:

This Research Report is available. This report has been submitted for publication outside of IBM and will probably be copyrighted if accepted for publication. It has been issued as a Research Report for early dissemination of its contents. In view of the transfer of copyright to the outside publisher, its distribution outside of IBM prior to publication should be limited to peer communications and specific requests. After outside publication, requests should be filled only by reprints or legally obtained copies of the article (e.g., payment of royalties). I have read and understand this notice and am a member of the scientific community outside or inside of IBM seeking a single copy only.

rc25353.pdf

Questions about this service can be mailed to reports@us.ibm.com .