A Comprehensive Analysis of Concurrency Control Performance for Centralized Databases

  • Werner Kiessling
  • Harald Pfeiffer

Abstract

The behavior of concurrency control methods for centralized databases is investigated and evaluated on the basis of discrete-event simulation. The gained results give new insights concerning the quantitative overhead introduced by such synchronization methods. The establishment of a uniform cost measurement model enables a qualitative assessment and comparison of several competing methods, the (r, x) — (r, a, x) —, (r, a, c) — lockprotocol and the optimistic method. In particular, valuable results concerning the effective parallelism, the tradeoff between early and late serialization, real time overhead for cycle searching and lengths of cycles are presented. The simulation series are driven by synthetic transaction workloads and also by a real reference string. Finally, the impact of these results on the selection of a suitable concurrency control method for practical applications is discussed.

Keywords

Expense Straw Dura Lost Tocol 

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. [1]
    Agrawal ; Corey; DeWitt: Deadlock Detection is Cheap, University of California, Berkeley, ERL Memorandum No. M83/5, Jan. 1983.Google Scholar
  2. [2]
    Bayer, R.: Database System Design for High Performance, IFEP 1983, Paris, pp. 147–155.Google Scholar
  3. [3]
    Bayer; Elhardt; Heigert; Reiser: Dynamic Timestamp Allocation for Transactions in Database Systems, in Distr. Data Bases, ed. H. J. Schneider, North- Holland, pp. 9–20, 1982.Google Scholar
  4. [4]
    Corey, MJ.: An Abstract Model of Database Concurrency Control Algorithms, SIGMOD 1983, pp. 97–107.Google Scholar
  5. [5]
    Gray; Homan; Obermarck; Korth: A Straw Man Analysis of Probability of Waiting and Deadlock IBM Res. Lab. San Jose, RJ3066(38112), 2126/81.Google Scholar
  6. [6]
    Kiessling, W Landherr, G.: A Quantitative Comparison of Lockprotocols for Centralized Databases, Proc. VILDB, Florence, 1983, pp. 120–130.Google Scholar
  7. [7]
    Kiessling, W Pfeiffer, H: A Comprehensive Analysis of Lock protocols for Centralized Database Systems, Research Report TUM-INFO 8402, Technical Univ. Munich, Febr.1984.Google Scholar
  8. [8]
    Kung, H.T Robinson, J.: On Optimistic Methods for Concurrency Control, Pros. VLDB, Rio de Janeiro, Oct 1979.Google Scholar
  9. [9]
    Lin, W. K.; Iblte, J.: Basic Timestamp, Multiple Version Timestamp, and Two-Phase Locking, Pros VLDB, Florence 1983, pp. 109–119.Google Scholar
  10. [10]
    Menasce, D.A Nakauishc, T.: Optimistic Versus Pessimistic Concurrency Control Mechanisms in Database Management Systems, in Inf. Syst, VoL 7, Na 1, 1982, pp. 13–27.CrossRefMATHGoogle Scholar
  11. [11]
    Peinl; Reuters Empirical Comparison of Database Concurrency Control Schemes, Pros. VLDB, Florence, 1983.Google Scholar
  12. [12]
    Potier, O; Leblanc, Ph.: Analysis of Locking Policies in Database Management Systems, in Comm. ACM Oct. 1980, Vol. 23, No. 10, pp. 584–593.CrossRefGoogle Scholar
  13. [13]
    Rosenkrantz; Steams; Lew is: System Level Concurrency Control for Distributed Database Systems, ACM TODS, Vol. 3, Na. 2, June 1978, pp. 178-198.CrossRefGoogle Scholar
  14. [14]
    Tay, Y-C; Suri, R.: Choice and Performance of Locking for Databases, V-LDB Singapore, Aug. 1984, pp. 199–127.Google Scholar
  15. [15]
    Tay, Y.C.; Goodman, N,: Suri, R.: Performance Evaluation of Locking in Databases: A Survey, Harvard Univ., TR-17-84, 1984.Google Scholar

Copyright information

© Springer-Verlag New York Inc. 1985

Authors and Affiliations

  • Werner Kiessling
    • 1
    • 2
  • Harald Pfeiffer
    • 2
  1. 1.University of CaliforniaBerkeleyUSA
  2. 2.Institut für InformatikTechnische Universität MünchenWest-Germany

Personalised recommendations