Performance tradeoffs in real-time transaction processing with client data caching

Abstract

The client-cached data can enhance performance of transaction processing in a client-server DBMS. It can improve the throughput of the committed transactions and can reduce the response time of each individual transaction. Does this client data caching also lend itself to processing of real-time transactions with timing constraints? In this paper, we investigate the performance tradeoffs in real-time transaction processing with client data caching in a client-server DBMS. Through a series of simulation experiments with two types of algorithms RTS/NC(Real-Time Scheduling with No Cache) and RTS/C(Real-Time Scheduling with Cache), we present and analyze how the performance of real-time transaction processing is affected when the number of clients, the client cache size, and the write probability of transactions vary.