File Replication for Enhancing the Availability of Parallel I/O Systems on Clusters

  • Authors:
  • Hau-Yang Cheng;Chung-Ta King

  • Affiliations:
  • -;-

  • Venue:
  • IWCC '99 Proceedings of the 1st IEEE Computer Society International Workshop on Cluster Computing
  • Year:
  • 1999

Quantified Score

Hi-index 0.00

Visualization

Abstract

Distributed environments such as networks of workstations are becoming more cost-effective and popular, and more high performance computations are moving into such environments. Although in the scientific computing arena NOWs have been used mainly for their high performance, their potential capability in reliability and high availability has not been fully exploited. One important area where this capability can be exploited on NOWs is to ensure data reliability and availability through the parallel I/O system. In this paper, we investigate the availability issues in parallel I/O systems with a shared-nothing disk configuration. A new file replication method, called locality aware file replication (LAFR), is proposed. LAFR is application specific and is designed specially for a parallel file with multiple access patterns. By comparing with previously proposed solutions, we demonstrate that the LAFR scheme may lead to better data access locality, while maintaining the same level of availability.