LISA 2002 - Technical Program Abstract
Geographically Distributed System for Catastrophic Recovery
Kevin Adams - NSWCDD
Pp. 47-64 of the Proceedings of LISA '02:
Sixteenth Systems Administration Conference,
USENIX Association, 2002).
This paper presents the results of a proof-of-concept
implementation of an on-going project to create a cost effective
method to provide geographic distribution of critical portions of a
data center along with methods to make the transition to these backup
services quick and accurate. The project emphasizes data integrity
over timeliness and prioritizes services to be offered at the remote
site. The paper explores the tradeoff of using some common clustering
techniques to distribute a backup system over a significant
geographical area by relaxing the timing requirements of the cluster
technologies at a cost of fidelity.
The trade-off is that the fail-over node is not suitable for high
availability use as some loss of data is expected and fail-over time
is measured in minutes not in seconds. Asynchronous mirroring,
exploitation of file commonality in file updates, IP Quality of
Service and network efficiency mechanisms are enabling technologies
used to provide a low bandwidth solution for the communications
requirements. Exploitation of file commonality in file updates
decreases the overall communications requirement. IP Quality of
Service mechanisms are used to guarantee a minimum available bandwidth
to ensure successful data updates. Traffic shaping in conjunction with
asynchronous mirroring is used to provide an efficient use of network
Traffic shaping allows a maximum bandwidth to be set minimizing
the impact on the existing infrastructure and provides a lower
requirement for a service level agreement if shared media is used. The
resulting disaster recovery site, allows off-line verification of
disaster recovery procedures and quick recovery times of critical data
center services that is more cost effective than a transactionally
aware replication of the data center and more comprehensive than a
commercial data replication solution used exclusively for data
vaulting. The paper concludes with a discussion of the empirical
results of a proof-of-concept implementation.
- View the full text of this paper in
PDF, and Postscript. Until November 2003, you will need your USENIX membership identification in order to access the full papers.
The Proceedings are published as a collective work, © 2002 by the USENIX Association. All Rights Reserved. Rights to individual papers remain with the author or the author's employer. Permission is granted for the noncommercial reproduction of the complete work for educational or research purposes. USENIX acknowledges all trademarks within this paper.
- If you need the latest Adobe Acrobat Reader, you can download it from Adobe's site.