Real-Time Concurrency Control and Transaction Scheduling

Abstract

Existing real-time concurrency control and scheduling techniques are inadequate for controlling the execution of transactions in a real-time database system. Some of these techniques place unnecessary restrictions of the concurrent execution of tasks in order to ensure they are executed correctly. Several of these mechanisms lack modularity--an important property in database concurrency control. Most of these real-time scheduling mechanisms lack the flexibility necessary to schedule both periodic and event-driven transactions to meet their timing constraints. This thesis examines the desired properties of concurrency control and the desired properties and priority management issues of real-time transaction scheduling. A survey of some existing real-time concurrency control and scheduling mechanisms are presented and evaluated. A new integrated real-time concurrency control and transaction scheduling mechanism, teamed the dataflow scheduler, is presented. This scheduling technique meets all of the desired properties of concurrency control and provides more flexibility than existing real-time techniques. A roll-forward approach to data item conflict resolution is used by the scheduler. Simulation results show that this method performs better in meeting transactions deadlines that restarting or rolling back a transaction.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jan 01, 1989
Accession Number
ADA218162

Entities

People

  • Ronnie Edge

Organizations

  • Air Force Institute of Technology

Tags

Communities of Interest

  • Space

DTIC Thesaurus Topics

  • Air Force
  • Business Administration
  • Classification
  • Computations
  • Computers
  • Data Sets
  • Engineers
  • Information Science
  • Maintenance
  • Maintenance Personnel
  • Operating Systems
  • Security
  • Simulations
  • Students
  • Systems Engineering
  • United States
  • Universities

Fields of Study

  • Computer science
  • Engineering

Readers

  • Parallel and Distributed Computing.