Comparison of File Systems
Finding an “optimal” file system for CAE applications — especially for equation solvers such as the direct sparse solver, which can be disk I/O-intensive — is not straightforward.
Here are some general benchmarks (I’m sure you can google a ton more):
- http://fsbench.netnation.com/
- http://linuxgazette.net/122/TWDT.html#piszcz
- http://www.debian-administration.org/article/Filesystems_ext3_reiser_xfs_jfs_comparison_on_Debian_Etch
The general consensus from above is that XFS or JFS are the best all-around file systems. (For ANSYS applications, the most important consideration may be file systems that provide good performance for reading/writing large files.)
A couple of years ago, I ran some ANSYS 11.0 benchmarks on three file systems: ext3, JFS, and ReiserFS. I only ran the benchmark once for each file system (since the full benchmark took a while to complete), and the results were generally within 5% of each other (with one or two anomalies), so the results were inconclusive. I think that part of the reason was that I was running on a RAID array (hardware RAID controller, not software RAID), so the disk I/O was fast, independent of the file system. (If I wanted to run an ‘actual’ benchmark, I would’ve run the benchmark a couple of times for each file system not using RAID, for example. However, I was just testing to see which file system would be best for my PC configuration.) As a result, if you have a fast RAID array (which ideally should be used for ANSYS applications), the file system may not play as critical a role as you may think, and other considerations may be used for the selection of a file system.
Comments are closed.