Workload Class Importance Policy in Autonomic Database Management Systems

  • Authors:
  • Harley Boughton;Pat Martin;Wendy Powley;Randy Horman

  • Affiliations:
  • Queen's University, Canada;Queen's University, Canada;Queen's University, Canada;IBM Toronto Lab, Canada

  • Venue:
  • POLICY '06 Proceedings of the Seventh IEEE International Workshop on Policies for Distributed Systems and Networks
  • Year:
  • 2006

Quantified Score

Hi-index 0.00

Visualization

Abstract

A key advantage of Autonomic Computing Systems will be their ability to manage according to business policies. A key challenge to realizing this ability is the problem of automatically translating high-level business policies into low-level system tuning policies, which is the result of the different semantics used at the two levels. Economic models, which are expressed using business level concepts, have been used successfully in computer resource allocation problems. In this paper, we utilize an economic model to map business policies to resource allocation decisions in a database management system (DBMS). We focus on business policies that describe the relative importance of competing workloads on a DBMS. We present experiments with a simulation of the model that investigate a number of meanings of importance and identify how this additional information can be used to effectively allocate main memory resources in a commercial DBMS.