Enterprise Integration & Modeling: Metadatabase Research Home
 home || MDB Research | Virtual Lab | Case Tool | Downloads | Publications | Researchers | Links

Differential Control on Distributed Database Updates Using Concurrent Rule-Based Shells

Abstract

Distributed database updates is a hard problem. A root cause to this problem is the fact that the traditional control method always requires full control (instantaneous correctness criterion) for all common/global data; which makes the global transaction serialization impractical in significantly distributed environments. The two-phase locking approach is subject to this performance problem, and so do other serializability relaxation approaches. In addition, the traditional approaches lack adaptiveness, i.e., the ability to dynamically adjust control requirements at the run time in order to either optimize or reduce the complexity as the real needs/usage of the database evolves. In this research, we propose a new differential concurrency control method using the usage correctness criterion to manage distributed database updates. This new method, based on the Metadatabase model, includes (1) an algorithm to determine the theoretical complete set of data updates rules and data integrity rules, (2) a rule-based representation method for usage correctness criteria to prioritize and reduce the control rules, (3) a new rule-based shell design to concurrently process the controls, and (4) a synchronizer to ensure the serializability under usage criteria and to improve the performance using the concurrent shells. The correctness and relevance of the new method are proven in both theoretical verification and empirical validation. The formal analysis establishes that the new method is correct and comparable to the traditional approaches under the same instantaneous correctness criterion; while it also holds promises for improved performance. More important is the fact that the new method is uniquely capable of managing directly and adaptively the control requirements beyond the design time correctness criterion, and thereby assures practicality for a control. The empirical validation includes a laboratory testing and an industrial case - the Product Data Management System at Samsung Electronic Corporation. The results of the laboratory testing (an LAN setting) confirms the theoretical analysis and substantiates the design claims on correctness, feasibility, and practicality. The Samsung case study (a WAN setting) further shows the industrial relevance of the usage correctness criterion and the scalability of the new method in a large scale industrial setting. A three-to-four orders of magnitude of performance improvement is recorded in the preliminary testings concerning some parts of the update process where the new method is directly employed.

 

viu.eng.rpi.edu is hosted by Professor Cheng Hsu.
Rensselaer Polytechnic Institute
Department of Industrial and Systems Engineering (formally Decision Sciences & Engineering Systems)
110 8th St., Center for Industrial Innovation, Room 5123, Troy, NY 12180-3590

Copyright 1997-2014. MetaWorld, Nothing on this site may be commercially used without written consent.

Valid XHTML 1.0! Valid CSS!