Skip to main content
U.S. Department of Energy
Office of Scientific and Technical Information

Performance of update algorithms for replicated data in a distributed database

Technical Report ·
OSTI ID:5702301
The performance of update algorithms for replicated data in a distributed database and several other related issues are studied. A simple model of a distributed database which is suitable for studying updates and concurrency control is presented. A performance model and a set of parameters which represent the most important performance features of a distributed database are developed. The distributed database models are used to study the performance of update algorithms for replicated data. This is done in two steps. First the algorithms are analyzed in the case of completely replicated databases in a no-failure, update-only environment. Then, the restrictions are eliminated one at a time, and the impact on the system performance of doing this is evaluated. The MCLA-h algorithm performs better than all other update algorithms in most cases of interest. The MCLA-h algorithm is based on the concept of hole lists. To investigate the validity of our results in a general system, the assumptions made initially in the performance studies are relaxed. It is shown that it is possible to make a centralized control algorithm (like the MCLA-h) resilient in the face of many types of failures. Distributed databases with partitioned data and multiple independent control mechanisms are analyzed. The performance results of the fully duplicated case can be used to design update algorithms in this environment. The processing of updates that do not specify their base set initially is demonstrated. 47 figures, 11 tables.
Research Organization:
Stanford Univ., CA (USA). Dept. of Computer Science
DOE Contract Number:
EY-76-C-03-0515
OSTI ID:
5702301
Report Number(s):
STAN-CS-79-744
Country of Publication:
United States
Language:
English