Using Periodic Patterns for Dynamic Estimation of Future Database Workloads

Abstract

Analysis of historical database workloads can be used to improve the future performance of database systems. This work investigates a problem how to use information about periodic processing of database applications recorded in the past for static and dynamic estimation of the future database workloads. The new concepts of anchor events and event expressions are used to create the mappings of past periodic patterns in data processing into the future database workloads. We show that event expressions allow for both static and dynamic estimation of future database workloads. The paper presents the algorithms for static a dynamic estimation of future database workloads from information about time distributions of periodic processing of database applications and information about the events that trigger processing of the applications.