A Hybrid Distributed Centralized System Structure for Transaction Processing

Abstract

In a fully centralized transaction processing system, all transaction input messages are shipped to the central system, where the transaction is processed, and output messages are sent back to the terminal; hence, the centralized system has this communications overhead and delay regardless of geographical locality of reference. On the other hand, the performance of a fully distributed system depends critically on the number of remote data accesses by a transaction; the performance of the distributed system is better than the centralized system if the number of remote data accesses per transaction is small, but is much worse otherwise. In this paper a hybrid system structure is examined that consists of distributed systems to take advantage of locality of reference, and a central system to handle transactions that access nonlocal data. We note that several transaction processing applications such as reservation systems, insurance, and banking have such regional locality of reference. A concurrency and coherency control protocol is described that maintains the integrity of the data and has good performance for transactions that access local or nonlocal data. It is shown that the performance of the hybrid system is much less sensitive to the fraction of remote accesses than the distributed system, and offers similar performance to the distributed system for local transactions. © 1990 IEEE