Distributed Database Performance Modeling

Abstract : This paper describes a distributed database model that simulates database utilization in the SPADOC data processing environment. The model is table driven such that database access requirements, file locations, and other information necessary for definig the database environment are set up internally in several tables. Each operation for the SPADOC mission is defined and represented by a transaction flow diagram (TFD) and a sequence of TFDs representing a mission is input to the model. The model executes input TFDS by looking up tables that specify their data file access requirements. While executing TFDs, the history of database usage is logged and various statistics are gathered. The performance data are used to measure database access overhead, characterize database workload, and fine-tune performance by reallocating files. The model has been implemented in Fortran on the VAX 11/780 and has been used for both measuring a given database and analyzing sensitivity to design changes. It has proven to be an effective tool for measuring and analyzing design effectiveness of distributed databases. (Author)