Automated Drill Down: An Approach to Automated Problem Isolation for Performance Management

Increasing the level of automation for performance management is essential if installations are to cope with the growing number and diversity of systems that they manage. Although much effort has been devoted to expert systems for performance management, these tools have not resulted in substantial productivity benefits. Herein, we take a different approach. We observe that a first step in problem determination is problem isolation-determining the general characteristics of the problem at hand. Recent developments in area of multidimensional databases. (MDDBs) can greatly aid in problem isolation by providing a way to "slice and dice" data in different ways. We propose a layer on top of MDDBs that provides for automated problem isolation. A prototype demonstrating this technology has been implemented in a commercial MDDB (Cognos Power Play).