Klaus Elhardt
Abstract
Performance in database systems is strongly influenced by buffer management and transaction recovery methods. This paper presents the principles of the database cache, which replaces the traditional buffer. In comparison to buffer management, cache management is more carefully coordinated with transaction management, and integrates transaction recovery. High throughput of
small- and medium-sized transactions is achieved by fast commit processing and low database traffic. Very fast handling of transaction failures and short restart time after system failure are guaranteed in such an environment. Very long retrieval and update transactions are also supported.
- 1 BAYER, R. Integrity, concurrency, and recovery in databases. In Proceedings 1st Conference of the EC1 (Amsterdam, 1976). Lecture Notes in Computer Science, 44. Springer Verlag, New York, 1976,79-106. Google Scholar
Digital Library
- 2 BAYER, R., HELLER, H., AND REISER, A. Parallelism and recovery in database systems. AC34 Trans. Database Syst. 5, 2 (June 1980), 139-156. Google ScholarDigital Library
- 3 BAYER, R., ELHARDT, K., HEIGERT, J., AND REISER, A. Dynamic timestamp allocation for transactions in database systems. In Proceedings 2nd International Symposium on Distributed Data Bases (Berlin, 1982), H.-J. Schneider, Ed., North-Holland, 1982, 9-20.Google Scholar
- 4 BELADY, L.A. A study of replacement algorithms for virtual storage computers. IBM Syst. J. 5, 2 (1966), 78-101.Google ScholarDigital Library
- 5 CHEW T.C. Computer technology and the database user. In Proceedings 4th International Conference on Very Large Data Bases (Berlin, 1978), 12-86.Google Scholar
- 6 ELHARDT, K. The database cache: Design principles, algorithms, properties. Dissertation, TUM- 18208,1982, Technical Univ. Munich (in German).Google Scholar
- 7 ESWARAN, K.P., GRAY, J. N., LORIE, R. A., AND TRAIGER, I. L. The notions of consistency and predictive locks in a database system. Commun. ACM 19, 11 (1976), 624-633. Google ScholarDigital Library
- 8 FOSSUM, B.M. Database integrity as provided for by a particular database management system. In Proceedings ZFZP Working Conference on Data Base Management, J.W. Klimbie and K.L. Koffeman, Eda., North-Holland, 1974,271-288.Google Scholar
- 9 GRAY, J.N., MCJONES, P., BLASGEN, M., LINDSAY, B., LORIE, R., PRICE, T., PUTZOLU, F., AND TRAICER, I. L. The recovery manager of the System R database manager. ACM Comput. Suru. 13,2 (1981), 223-242. Google ScholarDigital Library
- 10 GRAY, J.N. Notes on database operating systems. In Lecture Notes in Computer Science, 60, Springer-Verlag, New York, 1978,393-481. Google ScholarDigital Library
- 11 HAERDER, T., AND REUTER, A. Principles of transaction-oriented recovery-a taxonomy. Internal Rep. 50/82, Univ. Kaiserslautem, 1982.Google Scholar
- 12 HAERDER, T., AND REUTER, A. Optimization of logging and recovery in a database system. In Proceedings IFIP Working Conference on Data Base Architecture (Venice, 1979), G. Bracchi and G.M. Nijssen, Eds., North-Holland, 1979, 151-168.Google Scholar
- 13 HERRMANN, G. Measurement of the behavior of the database cache by application of a page reference string. Diplomarbeit, Technical Univ. Munich, 1983 (in German).Google Scholar
- 14 ZMS/VS Version 1, FastPath Feature, General Information Manual. GH20-9069-2.Google Scholar
- 15 KING, W.F. Relational database systems: Where we stand today. In Proceedings ZFZP Congress on Information Processing, S. Lavington, Ed., North-Holland, 1980. Google ScholarDigital Library
- 16 KUNG, H.T., AND ROBINSON, J.T. On optimistic methods for concurrency control. ACM Trans. Database Syst. 6, 2 (1981), 213-226. Google ScholarDigital Library
- 17 LINDSAY, B., SELINGER, P., GALTIERI, C., GRAY, J., LORIE, R., PRICE, T., PUTZOLU, F., TRAIGER, I.L., AND WADE, B. Notes on distributed databases. IBM Rep. RJ2571, July 1979.Google Scholar
- 18 LORIE, R.A. Physical integrity in a large segmented database. ACM Trans. Database Syst. 2, 1 (1977), 91-104. Google ScholarDigital Library
- 19 MCGEE, W.C. The information management system IMS/VS. IBM Syst. J. 16, 2 (1977), 84- 168.Google Scholar
- 20 REUTER, A. A fast transaction-oriented logging scheme for UNDO-recovery. IEEE Trans. Softw. Eng. SE-6 (1980), 348-356.Google ScholarDigital Library
- 21 SIWIEC, J.E. A high-performance DB/DC system. IBM Syst. J. 16,2 (1977), 169-195.Google ScholarDigital Library
- 22 Software Product CJDS, Universal Database Management System. UDS-V2 Reference Manual Package, Siemens AG, Munich.Google Scholar
- 23 SQL/DATA SYSTEM, Application Programmer's Guide. Form No. Sh24-5018-1, IBM, 1982.Google Scholar
- 24 SQL/DATA SYSTEM, Terminal User's Reference. Form No. Sh24-5016, IBM, 1982.Google Scholar
- 25 TANDEM Computers Inc., Systems Introduction. 1975.Google Scholar
- 26 YAMAMOTO, T., AND KAMOGAWA, K. The seat reservation system in Japanese national railways. Information Systems Management Center JNR. Unpublished paper.Google Scholar
Index Terms
- A database cache for high performance and fast restart in database systems
Recommendations
Fast Serializable Multi-Version Concurrency Control for Main-Memory Database Systems
SIGMOD '15: Proceedings of the 2015 ACM SIGMOD International Conference on Management of DataMulti-Version Concurrency Control (MVCC) is a widely employed concurrency control mechanism, as it allows for execution modes where readers never block writers. However, most systems implement only snapshot isolation (SI) instead of full ...
Reviews
Vijay Varadachari Raghavan
There is a need to have systems which achieve extremely high throughput of short transactions. At present, systems that can meet this need are special-purpose systems since commercially available full-function DBMSs are not fast enough. This paper proposes an approach, involving the use of DB cache to replace the conventional buffer, which can enable full-functional systems to provide the throughput and short restart times necessary. This is considered feasible as a consequence of the recent hardware advances that have led to enormous increase in main memory available with large mainframe. DB cache is a large main memory space (volatile). Safe is the other component needed; it is a nonvolatile memory with fast sequential access, and it is about the size of the cache. The cache acts as a very large buffer for DB pages, but copies (dirty pages) are never swapped-out from the cache to the physical DB. In other words, updates of a transaction are kept in the cache until the transaction commits; thus, the physical DB pages are always clean. The version of a page in DB may be obsolete, in which case the cache holds the valid version. Safe backs up the cache; its aim is to be able to reconstruct any valid pages of the cache, not yet written to the physical DB, which would be lost in the event of a system failure. The DB cache method is shown to have the following advantages: :9BEfficient page addressing schemes can be used. Frequently used DB pages are managed without DB traffic. There is no need for the logging of before images. Fast commit processing is achieved. Transaction failures require no I/O. Restart after system failure is handled without DB traffic. The implications of this scheme for media failures and a way of dealing with long transactions without hampering performance are also discussed. Simulation results are provided to obtain some insight on how to determine the size of the DB cache needed in a given situation.
Access critical reviews of Computing literature here
Become a reviewer for Computing Reviews.
Comments