Adaptive Data Stream Management System Using Learning Automata

TL;DR

This paper proposes an adaptive Data Stream Management System that uses Learning Automata to dynamically tune parameters, significantly improving response time and system adaptivity in unpredictable data stream environments.

Contribution

It introduces a novel architecture integrating Learning Automata for parameter tuning to enhance DSMS adaptivity and response time.

Findings

Parameters reach optimal values over time.

System adaptivity improves significantly.

Response time is reduced after tuning.

Abstract

In many modern applications, data are received as infinite, rapid, unpredictable and time- variant data elements that are known as data streams. Systems which are able to process data streams with such properties are called Data Stream Management Systems (DSMS). Due to the unpredictable and time- variant properties of data streams as well as system, adaptivity of the DSMS is a major requirement for each DSMS. Accordingly, determining parameters which are effective on the most important performance metric of a DSMS (i.e., response time) and analysing them will affect on designing an adaptive DSMS. In this paper, effective parameters on response time of DSMS are studied and analysed and a solution is proposed for DSMSs' adaptivity. The proposed adaptive DSMS architecture includes a learning unit that frequently evaluates system to adjust the optimal value for each of tuneable effective. Learning Automata is used as the learning mechanism of the learning unit to adjust the value of tuneable effective parameters. So, when system faces some changes, the learning unit increases performance by tuning each of tuneable effective parameters to its optimum value. Evaluation results illustrate that after a while, parameters reach their optimum value and then DSMS's adaptivity will be improved considerably.