There has been increasing research effort in applying control-theoretic approaches to performance management for computer systems such as Internet web servers, databases and storage systems. Since today's Internet servers and applications are often operated under dynamically changing load conditions, linear control designs may not suffice to provide desired performance guarantees. This motivates nonlinear system modeling and control design methodologies. This paper studies the admission control for an Internet web server. It presents a linear-parameter-varying (LPV) approximation for the modeling of the dynamic relationship from the request rejection ratio to the response time for the admitted requests. The time-varying workload parameter, in particular the workload intensity, is specified as the scheduling variable that is used to parameterize the LPV model. An LPV system identification algorithm is applied to derive the empirical model, and then an LPV-H∞ controller is designed to provide response time guarantees. The performance of the resulting LPV control compares favorably to that of a linear design. The utilization of scheduling parameters can be generalized to accommodate more sophisticated workload characterizations and more complicated server environments. By exploring the nature of dependence of server performance on time-varying load and operating conditions, the proposed general framework is possibly applicable to a diverse spectrum of server-based applications.
All Science Journal Classification (ASJC) codes
- Control and Systems Engineering
- Computer Science Applications
- Applied Mathematics
- Electrical and Electronic Engineering