Susan B. Davidson
Abstract
A protocol for transaction processing during partition failures is presented which guarantees mutual consistency between copies of data-items after repair is completed. The protocol is “optimistic” in that transactions are processed without restrictions during failure; conflicts are then detected at repair time using a precedence graph, and are resolved by backing out transactions according to some backout strategy. The resulting database state then corresponds to a serial execution of some subset of transactions run during the failure. Results from simulation and probabilistic modeling show that the optimistic protocol is a reasonable alternative in many cases. Conditions under which the protocol performs well are noted, and suggestions are made as to how performance can be improved. In particular, a backout strategy is presented which takes into account individual transaction costs and attempts to minimize total backout cost. Although the problem of choosing transactions to minimize total backout cost is, in general, NP-complete, the backout strategy is efficient and produces very good results.
- 1 AHO, A.V., HOPCROFT, J.E., AND ULLMAN, J.D. The Design and Analysis of Computer Algorithms. Addison-Wesley, Reading, Mass., 1974. Google Scholar
Digital Library
- 2 ALSBERG, P.A., AND DAY, J.D. A principle for resilient sharing of distributed resources. In Proceedings 2nd International Conference on Software Engineering {Oct. 13-15, 1976), 562-570. Google ScholarDigital Library
- 3 BERNSTEIN, P.A., ROTHNIE, J.B., GOODMAN, N., AND PAPADIMITRIOU, C.A. The concurrency control mechanism of SDD-I: A system for distributed databases (the fully redundant case). IEEE Trans. So{tw. Eng. 4, 3 (May 1978), 154-167.Google Scholar
- 4 BERNSTEIN, P.A., SHIPMAN, D.W., AND WONG, W.S. Formal aspects of serializability in database concurrency control. IEEE Trans. Softw. Eng. 5, 3 (May 1979), 203-216.Google ScholarDigital Library
- 5 DAVIDSON, S.B., GARCIA-MOLINA, H., AND SKEEN, D. Consistency in the face of partition failures: A survey. Tech. Rep. TR#84-4, Dept. of Computer and Information Sciences, Univ. of Pennsylvania, Feb. 1984.Google Scholar
- 6 DAVIDSON, S.B. An optimistic protocol for partitioned distributed database systems. Ph.D. dissertation, Dept. of Electrical Engineering and Computer Science, Princeton Univ. Princeton, N.J., Oct. 1982. Google ScholarDigital Library
- 7 ESWARAN, K.P., GARY, J.N., LOR{E, R.A., AND TRAIGER, I.L. The notions of consistency and predicate locks in a database system. Commun. ACM I9, 11 (Nov. 1976), 624-633. Google ScholarDigital Library
- 8 GAREY, M.R., AND JOHNSON, D.S. Computers and Intractability: A Guide to the Theory of NP- Completeness. W.H. Freeman and Co., San Francisco, 1979. Google ScholarDigital Library
- 9 GARCIA-MOLINA, H. Elections in a distributed computing system. IEEE Trans. Comput. C-31, 1 (Jan. 1982), 48-59.Google ScholarDigital Library
- 10 GARCIA-MOLINA, H. Using semantic knowledge for transaction processing in a distributed database. ACM Trans. Database Syst. 8, 2 {June 1983), 186-213. Google ScholarDigital Library
- 11 GIFFORD, D.K. Weighted voting for replicated data. In Proceedings 7th Symposium on Operating System Principles (Pacific Grove, Calif., Dec. 10-12, 1979), ACM, New York, 150-162. Google ScholarDigital Library
- 12 HAMMER, M.M., AND SHIPMAN, D.W. The reliability mechanisms of SDD-I: A system for distributed databases. ACM Trans. Database Syst. 5, 4 (Dec. 1980), 431-466. Google ScholarDigital Library
- 13 HOPCROFT, J.E., AND KARP, R.M. An n25 algorithm for maximum matching in bipartite graphs. J. SIAM Comput 2 (1973), 225-231.Google ScholarCross Ref
- 14 JOHNSON, D.B. Finding all the elementary circuits of a directed graph. SIAM J. Comput. 4 (1975), 77-84.Google ScholarDigital Library
- 15 KARP, R.M. Reducibility among combinatorial problems. In Complexity of Computer Computations, R.E. Miller and J.W. Thatcher, Eds., Plenum Press, New York, 85-103.Google Scholar
- 16 LAWLER, E.L. Combinatorial Optimization: Networks and Matroids. Holt, Rinehart and Winston, New York, 1976.Google Scholar
- 17 LIPSKI, W. On semantic issues connected with incomplete information databases. A CM Trans. Database Syst. 4, 3 (Sept. 1979), 262-296. Google ScholarDigital Library
- 18 LYHCH, N.A. Multilevel atomicity--a new correctness criterion for distributed databases. ACM Trans. Database Syst. 8, 4 (Dec. 1983), 484-502. Google ScholarDigital Library
- 19 MINOtJRA, T., AND WEIDERHOLD, G. Resilient extended true-copy token scheme for a distributed database system. IEEE Trans. Softw. Eng. SE8, 3 (May 1982), 173-189.Google Scholar
- 20 PAPADIMITRIOU, C.H. The serializability of concurrent database updates. J. ACM 26, 4 (Oct. 1979), 631-653. Google ScholarDigital Library
- 21 PAPADIMITRIOU, C.H., AND STEIGLITZ, K. Combinatorhl Optimization: Algorithms and Complexity. Prentice Hall, Englewood Cliffs, N.J., 1982. Google ScholarDigital Library
- 22 PARKER, D.S., POPEK, G.J., RUDISIN, G., STOUGHTON, A., WALKER, B., WALTON, E., CHOW, J., EDWARDS, D., KISER, S., AND KLINE, C. Detection of mutual inconsistency in distributed systems. In Proceedings 5th Berkeley Workshop on Distributed Data Management and Computer Networks (Feb. 1981), 172-183.Google Scholar
- 23 PARKER, D.S., AND RAMOS, R.A. A distributed file system architecture supporting high availability. In Proceedings 6th Berkeley Workshop on Distributed Data Management and Computer Networks (1982), 161-183.Google Scholar
- 24 REINGOLD, E.M., NEIVERGELT, J., AND DEO, N. CombinatorialAlgorithms: Theory and Practice. Prentice-Hall, Englewood Cliffs, N.J., 1977, 348-352. Google ScholarDigital Library
- 25 ROTHNIE, J.B., AND GOODMAN, N. A survey of research and development in distributed database management. In Proceedings 3rd International Conference on Very Large Databases (Tokyo, Oct. 1977), 48-61.Google Scholar
- 26 STEARNS, R.E., LEWIS, P.M., II, AND ROSENKRANTZ, D.J. Concurrency control for database systems. In Foundations of Computer Science (1976), 19-32.Google Scholar
- 27 THOMAS, R.H. A solution to the concurrency control problem for multiple copy databases. In IEEE Compcon {Spring 1978), 56-62.Google Scholar
- 28 TRAIGER, I.L., GRAY, J.N., GALTIERI, C.A., AND LINDSAY, B.G. Transactions and consistency in distributed database systems. IBM Res. Rep. RJ2555 33155, June 5, 1979.Google Scholar
- 29 WRIGHT, D.D. Merging partitioned databases. Ph.D. dissertation, TR83-575, Dept. of Computer Science, Cornell Univ., Ithaca, N.Y., Sept. 1983. Google ScholarDigital Library
Index Terms
- Optimism and consistency in partitioned distributed database systems
Recommendations
Reviews
Jason Gait
This paper describes a protocol for recovering from partitions in fully replicated distributed database systems. The protocol is optimistic in allowing transactions to proceed to their commit point during the partition, and only guarantees globally consistent data after the partition is recovered. The author proposes a method for automating conflict detection and transaction backout, based on precedence graphs whose cycles correspond to conflicts. When a partition is rejoined, a coordinating site is designated in each half of the partition, and all processing of transactions is temporarily blocked. Each coordinator constructs a totally ordered transaction history for its half of the partition, and the tansaction histories are merged to form the precedence graph. The edges in the precedence graph represent the required ordering of transactions, with ripple edges representing the effect of backing out transactions and interference edges representing conflicts between transactions in different partition halves. A backout strategy is employed to select some set of transactions to be backed out until the precedence graph is acyclic. The partitions can then be merged to form a mutually consistent global transaction history. Backout strategies attempt to minimize the cost of backed out transactions, but the problem of selecting a minimal cost set of transactions to backout is NP-complete. The paper presents simulation results for several heuristic backout strategies. The most successful backout strategy detects and breaks two-cycles before looking for longer cycles. The simulations indicate there is on the average just one long cycle remaining after removal of the two-cycles. The problem of breaking the two-cycles for th case of two partition groups is polynomial in the total number of nodes, and if no long cycles remain after breaking the two-cycles, then minimizing backout cost is also a polynomial problem.
Access critical reviews of Computing literature here
Become a reviewer for Computing Reviews.
Comments