Archive for the ‘parallel file systems’ Category

Another Improvement in Hard Drive Densities

Sunday, November 28th, 2010

At SC10 last week someone asserted that hard drives were a dead technology, and that attempting to improve the performance of spinning disks access was a waste of time. Galen was with me and together we pointed out that hard drives are going to be with us for the foreseeable future.

The belief that the end of disk density improvements is imminent comes up over and over with people who do not work in the storage area — but I’m here to tell you that HDD research and development is still occurring at a staggering rate. In the last several years, the rate of improvement for disk density has been increasing rather than decreasing. Perpendicular recording was a big deal. Sticking the bits on their end took us to 3TB of storage in 3.5″ drives.

Currently we are looking forward to three more improvements: patterned media, heat assisted magnetic recording, and shingling. And patterned media looks to be available soon. From the article we can expect this technology to drive our disks over the 10TB capacity level — which is a staggering amount of storage. With a little hand waving and projection its not hard to imagine HAMR and shingling to drive disk capacities close to 100TB. If we estimate half that for a 2.5″ drive, it still only takes 20 such drives to construct a Petabyte file system! All of these technologies are likely to be available at the time of Exascale machine construction, so I expect that hard disk drives are likely to play a significant role in the construction of Exascale storage systems — though I also expect other storage media (e.g. storage-class memory) will also be a significant element of future extreme storage systems.

A Study of Client-based Caching for Parallel I/O

Sunday, August 30th, 2009

Bradley W. Settlemyer.
“A Study of Client-based Caching for Parallel I/O.” [pdf]
Clemson University Doctoral Dissertation, Aug 2009.
(more…)

A Mechanism for Scalable Redundancy in Parallel File Systems

Wednesday, June 24th, 2009

Bradley W. Settlemyer.
“A Mechanism for Scalable Redundancy in Parallel File Systems.” [pdf]
Clemson University Master’s Thesis, May 2006.
(more…)

Using Server to Server Communication in Parallel File Systems to Simplify Consistency and Improve Performance

Wednesday, June 24th, 2009

Philip H. Carns, Bradley W. Settlemyer, and Water B. Ligon, III.
“Using Server to Server Communication in Parallel File Systems to Simplify Consistency and Improve Performance.” [pdf]
In The International Conference for High Performance Computing, Networking, Storage, and Analysis (SC ’08).
[21% acceptance, 59/277]
(more…)

A Technique for Lock-less Mirroring in Parallel File Systems

Tuesday, June 23rd, 2009

Bradley W. Settlemyer and Water B. Ligon, III.
“A Technique for Lock-less Mirroring in Parallel File Systems.”[pdf]
In The Workshop On Resiliency in High-Performance Computing (Resilience 2008)
at The 8th IEEE International Symposium on Cluster Computing and the Grid (CCGrid 2008).

(more…)