For application programs that access large, complex databases, performance tuning can be critical. This is particularly true for programs with stringent timing requirements. Furthermore, effective performance tuning can enable larger, more complex problems to be solved. Unfortunately, although a variety of automated tools have been developed that can be used to fine-tune performance, the overall tuning process can still be manually intensive and time consuming. In this talk, an agent-based prototype called DB-Assist will be described that automates many aspects of the tuning process. In particular, this talk will focus on one aspect of DB-Assist that deals with the general problem of object placement, where the goal is to distribute database objects to system devices so as to optimize performance. DB-Assist currently solves this problem by employing several heuristics. These heuristics will be presented, in addition to corresponding experimental results showing that for many realistic scenarios, the object placement problem can be solved both efficiently and optimally.
About the Speaker
Dr. Bernhard has 9 years of experience in the field of Computer Science. During that time he has specialized in database performance tuning, user interfaces, protocol/software testing, software engineering, and interconnection networks for parallel processing.
Current, he is a software engineer at Harris Information Systems Division (ISD), where he has worked on a variety of projects. At the present time he is the principal investigator on the IA4DB program for Rome Laboratory. The objective of this research is to develop a system for automatically fine-tuning the performance of Sybase relational databases. Other programs include the ArtX program - this involved the development of a natural language understanding system for processing text messages; and the Core program - this involved the re-development of NASAs space shuttle and space station test, checkout, control, and monitor systems.
Prior to joining Harris, he was a senior instructor and consultant for P.P. Texel of Eatontown, NJ. During that time his duties including consulting on Ada projects, and providing instruction in object-oriented analysis, object-oriented design, and Ada coding. Prior to joining Texel, he was a faculty member in the Computer Science Department at Clemson University in South Carolina. Courses taught included Software Engineering, Programming Methodology, Data Structures and Algorithms, Computational Complexity, and Theory of Computation.