Regulating Response Time in an Autonomic Computing System: A Comparision of Proportional Control and Fuzzy Control Approaches

TL;DR

This paper compares proportional and fuzzy control methods for dynamically tuning response time in an ecommerce Autonomic Computing system, demonstrating self-optimization in adjusting server parameters amid workload changes.

Contribution

It introduces and evaluates two control approaches for automating server parameter tuning to maintain response time in dynamic ecommerce environments.

Findings

Fuzzy control adapts better to workload variations.

Proportional control is simpler but less flexible.

Both methods improve response time regulation.

Abstract

Ecommerce is an area where an Autonomic Computing system could be very effectively deployed. Ecommerce has created demand for high quality information technology services and businesses are seeking quality of service guarantees from their service providers. These guarantees are expressed as part of service level agreements. Properly adjusting tuning parameters for enforcement of the service level agreement is time-consuming and skills-intensive. Moreover, in case of changes to the workload, the setting of the parameters may no longer be optimum. In an ecommerce system, where the workload changes frequently, there is a need to update the parameters at regular intervals. This paper describes two approaches, one, using a proportional controller and two, using a fuzzy controller, to automate the tuning of MaxClients parameter of Apache web server based on the required response time and the current workload. This is an illustration of the self-optimizing characteristic of an autonomic computing system.