RobuSTore: Robust Performance for Distributed Storage Systems

Huaxia Xia and Andrew A. Chien
CS2006-0851
March 13, 2006

Emerging large-scale scientific applications involve massive, distributed, shared data collections (petabytes), and require robust, high performance for read-dominated workloads. Achieving robust performance (low variability) in storage systems is difficult. We propose RobuSTore, a novel storage technique, which combines erasure codes and speculative access to reduce performance variability and increase performance. RobuSTore uses erasure codes to add flexible redundancy then spreads the encoded data across a large number of disks. Speculative access to the redundant data from multiple disks enables application requests to be satisfied with only early-arriving blocks, reducing performance dependence on the behavior of individual disks. We present the design and an evaluation of RobuSTore which shows improved robustness, reducing the standard deviation of access latencies by as much as 5-fold vs. traditional RAID. In addition, RobuSTore improves access bandwidth by as much as 15-fold. A typical 1 GB read from 64 disks has average latency of 2 seconds with standard deviation of only 0.5 seconds or 25%. RobuSTore secures these benefits at the cost of a 2-3x storage capacity overhead and ~1.5x network and disk I/O overhead.


How to view this document


The authors of these documents have submitted their reports to this technical report series for the purpose of non-commercial dissemination of scientific work. The reports are copyrighted by the authors, and their existence in electronic format does not imply that the authors have relinquished any rights. You may copy a report for scholarly, non-commercial purposes, such as research or instruction, provided that you agree to respect the author's copyright. For information concerning the use of this document for other than research or instructional purposes, contact the authors. Other information concerning this technical report series can be obtained from the Computer Science and Engineering Department at the University of California at San Diego, techreports@cs.ucsd.edu.


[ Search ]


NCSTRL
This server operates at UCSD Computer Science and Engineering.
Send email to webmaster@cs.ucsd.edu