Self-Organizing Storage Project Home Page
Quick Links
Overview
The goal of the SOS project is to build storage systems that tune
themselves to their workload. Our current focus is on general-purpose
network storage servers, because there has already been considerable
work in self-tuning storage systems for personal workstations and PCs.
SOS is one of the projects in the
Systems
Research at Harvard (SyRaH) group.
Project Status
The first phase of the project was to gather and analyze contemporary
NFS traces, in order to learn about the kinds of workloads that we
might expect to see on modern systems. We have gathered long-term
traces from three production environments: a CS research workload, an
email workload, and a general engineering/email workload. All of
these traces are available, in anonymized form, to any interested
researchers. We are also gathering whatever traces we can find, for
our own analysis and also to provide a public repository for research
data.
Visit our trace page for more information.
Publications and Tech Reports
- Ningning Zhu, Jiawu Chen, Tzi-cker Chiueh, and Daniel Ellard.
TBBT: Scalable and Accurate Trace Replay for
File Server Evaluation.
(PDF)
To be presented at
ACM SIGMETRICS'05,
June, 2005.
- Daniel Ellard.
Trace-Based Analyses and Optimizations for Network
Storage Servers.
(PDF)
PhD thesis,
Harvard
Computer Science Technical Report TR-11-04,
May 2004.
- Michael Mesnier, Eno Thereska, Daniel Ellard, Gregory R. Ganger and
Margo Seltzer.
File classification in self-* storage systems.
(PDF,
Postscript)
Carnegie
Mellon University Technical Report CMU-PDL-04-101,
January 2004.
To appear in the Proceedings of the
International
Conference on Autonomic Computing (ICAC-04).
- Daniel Ellard, Michael Mesnier, Eno Thereska, Gregory R. Ganger, and
Margo Seltzer.
Attribute-Based Prediction of File Properties.
(PDF,
Postscript)
Harvard
Computer Science Technical Report TR-14-03,
December, 2003.
- Ningning Zhu, Jiawu Chen, Tzi-cker Chiueh, and Daniel Ellard.
An NFS Trace Player for File System Evaluation.
(PDF,
Postscript)
Harvard
Computer Science Technical Report TR-16-03,
December, 2003.
- Daniel Ellard and Margo Seltzer.
New NFS Tracing Tools and Techniques for System Analysis
(PDF,
Postscript)
Appears in Proceedings of the 17th annual
Large Installation System Administration Conference
(LISA'03),
pages 73-85,
San Diego, CA.
October 2003.
(note that the page numbers in the on-line
version are not identical to the printed proceedings)
- Daniel Ellard and Margo Seltzer.
NFS Tricks and Benchmarking Traps
(PDF,
HTML)
Appears in
Proceedings of the
USENIX Annual Technical Conference, FREENIX Track,
pages 101-114,
San Antonio, Texas.
June 2003.
- Source code
for the changes described in the paper.
- Slides from the presentation
(PDF,
HTML).
- Errata: there is a typo in Figure 8 of the
published paper. The caption should read ``...
scsi1 runs 60-70% faster for all tests when
cursors are enabled. ide1 is only 50%
faster for the 2-stride test...'' This on-line
version of the paper is correct.
- Daniel Ellard, Jonathan Ledlie and Margo Seltzer.
The Utility of File Names
(PDF,
Postscript)
Harvard
Computer Science Technical Report TR-05-03,
March, 2003.
- Daniel Ellard, Jonathan Ledlie, Pia Malkani and Margo Seltzer.
Passive NFS Tracing
of Email and Research Workloads
(PDF,
HTML)
Appears in
Proceedings of the Second Annual
USENIX File
and Storage Technologies Conference (FAST'03),
pages 203-216,
San Francisco, California. March, 2003.
- Slides from the presentation
(PDF,
HTML).
- Daniel Ellard, Jonathan Ledlie, Pia Malkani and Margo Seltzer.
Everything You Always Wanted
to Know About NFS Trace Analysis, But Were Afraid to Ask
(PDF,
Postscript)
Harvard
Computer Science Technical Report TR-06-02,
June, 2002.
- Daniel Ellard.
Slides from the Birds-of-a-Feather session on File System Tracing
(PDF,
PostScript)
First USENIX File
and Storage Technologies Conference (FAST'02),
Monterey, California. January, 2002.
The SOS Project
<sos@eecs.harvard.edu>