A First-Principles Based Lpv Modeling And Design For Performance Management Of Internet Web Servers

This paper presents a control-theoretic approach to performance management of Internet Web servers to meet Service Level Agreements (SLA). In particular, a CPU frequency management problem is studied to provide response time guarantees with minimal energy cost. A linear uncertain model and a Linear-Parameter-Varying (LPV) system are derived based on first-principles analysis of transient and steady-state queueing dynamics from the allocated CPU resource to request response time. The LPV modeling utilizes the workload arrival and service parameters as scheduling variables, which allows the Web server to meet the response time SLA in the presence of dynamically changing load conditions. Using real Web server workloads, the performance of an LPV- H-infinity controller is compared to that of a linear controller designed at the high-percentile load parameters and a G/G/1-queueing based nonlinear optimization. The proposed LPV modeling and control framework can be generalized to incorporate more sophisticated workload models and more complicated server environments. In addition, due to the LPV nature of Web systems with respect to load conditions, the proposed approach can be applied to a variety of resource management problems and used for middleware design.