Concurrency control in computer services using adaptive optimal control

  • Authors:
  • Magnus Karlsson

  • Affiliations:
  • HP Laboratories, Palo Alto, CA

  • Venue:
  • MIC'06 Proceedings of the 25th IASTED international conference on Modeling, indentification, and control
  • Year:
  • 2006

Quantified Score

Hi-index 0.00

Visualization

Abstract

A shared computer service needs to be able to provide performance differentiation between workloads that are using the service so that their individual performance goals are met. A non-intrusive approach for this is to interpose a fair-queuing scheduler on the request path between clients and the service. One key problem in this context is to set the number of concurrent requests in the system so that the total system throughput is close to maximum while still providing performance differentiation. In this paper, we use adaptive optimal control to solve the problem and experimentally evaluate the control design on a real three-tier e-commerce site. The results show that the controller can successfully achieve both high throughput and good performance differentiation.