An Adaptive Data-Shipping Architecture for Client Caching Data Management Systems

  • Authors:

  • Kaladhar Voruganti;M. Tamer Özsu;Ronald C. Unrau

  • Affiliations:

  • IBM Almaden Research Laboratory, San Jose, USA. kaladhar@us.ibm.com;University of Waterloo, School of Computer Science, Waterloo, Canada. tozsu@uwaterloo.ca;YottaYotta, Edmonton, Canada. rcu@yottayotta.com

  • Venue:
  • Distributed and Parallel Databases
  • Year:
  • 2004

Quantified Score

Hi-index 0.00

Visualization

Abstract

Data-shipping is an important form of data distribution architecture where data objects are retrieved from the server, and are cached and operated upon at the client nodes. This architecture reduces network latency and increases resource utilization at the client. Object database management systems (ODBMS), file-systems, mobile data management systems, multi-tiered Web-server systems and hybrid query-shipping/data-shipping architectures all use some variant of the data-shipping. Despite a decade of research, there is still a lack of consensus amongst the proponents of ODBMSs as to the type of data shipping architectures and algorithms that should be used. The absence of both robust (with respect to performance) algorithms, and a comprehensive performance study comparing the competing algorithms are the key reasons for this lack of agreement. In this paper we address both of these problems. We first present an adaptive data-shipping architecture which utilizes adaptive data transfer, cache consistency and recovery algorithms to improve the robustness (with respect to performance) of a data-shipping ODBMS. We then present a comprehensive performance study which evaluates the competing client-server architectures and algorithms. The study verifies the robustness of the new adaptive data-shipping architecture, provides new insights into the performance of the different competing algorithms, and helps to overturn some existing notions about some of the algorithms.