September 2011

Conference Paper

Oak Ridge Leadership Computing Facility Position Paper

By:
Oral, Hakki S; Hill, Jason J; Thach, Kevin G; Podhorszki, Norbert ; Klasky, Scott A; Rogers II, James H; Shipman, Galen M
Publication Date:
September 9, 2011
Conference Name:
Fifth U.S. Department of Energy Best Practices Workshop on File Systems & Archives
Conference Location:
San Francisco, California, United States of America
Conference Sponsor:
DOE

Abstract

This paper discusses the business, administration, reliability, and usability aspects of storage systems at the Oak Ridge Leadership Computing Facility (OLCF). The OLCF has developed key competencies in architecting and administration of large-scale Lustre deployments as well as HPSS archival systems. Additionally as these systems are architected, deployed, and expanded over time reliability and availability factors are a primary driver. This paper focuses on the implementation of the “Spider” parallel Lustre file system as well as the implementation of the HPSS archive at the OLCF.