Bounded Inconsistency for Type-Specific Concurrency Control

  • Authors:

  • Man Hon Wong;Divyakant Agrawal;Hang Kwong Mak

  • Affiliations:

  • Dept. of Computer Science, The Chinese University of HK, Shatin, N.T., Hong Kong/ E-mail: mhwong@cs.cuhk.edu.hk;Dept. of Computer Science, University of California, Santa Barbara, CA 93106, USA/ E-mail: agrawal@cs.ucsb.edu;Dept. of Computer Science, University of California, Santa Barbara, CA 93106, USA/ E-mail: agrawal@cs.ucsb.edu

  • Venue:
  • Distributed and Parallel Databases
  • Year:
  • 1997

Quantified Score

Hi-index 0.00

Visualization

Abstract

The traditional correctness criterion of serializability indatabases is considered too restrictive especially whendatabases are used to model advanced applications. In general,two approaches are adopted to address this problem. The firstapproach considers placing more structure on data objects toexploit type specific properties while keeping serializabilityas the correctness criterion. The other approach uses explicitsemantics of transactions and databases to permit interleavedexecutions of transactions that are non-serializable. In thispaper, we attempt to bridge the gap between the two approachesby using the notion of serializability with boundedinconsistency. Users are free to specify the maximum level ofinconsistency that can be allowed in the executions ofoperations dynamically. In particular, if no inconsistency isallowed in the execution of any operation, the protocol will bereduced to a standard strict two phase locking protocol based ontype-specific semantics of data objects. On the other hand, ifinconsistency is not bounded, the execution of transactions isunrestricted in the proposed model. The proposed protocols havebeen implemented and the paper includes a performance comparisonwith the two-phase locking protocol. The results demonstrate thatthe associated overhead in the proposed protocol is notoverwhelming and the gains in transition throughput can besignificant for object based systems. Bounded inconsistency canbe applied to many areas which do not require exact values ofthe data such as for gathering information for statisticalpurpose, for decision support systems, and for reasoning inexpert systems which can tolerate uncertainty in input data.