[Beowulf] file IO benchmark
Many of your questions may have already been answered in earlier discussions or in the FAQ. The search results page will indicate current discussions as well as past list serves, articles, and papers.
Joachim Worringen joachim at ccrl-nece.deThu Nov 24 08:21:02 PST 2005
- Previous message: [Beowulf] file IO benchmark
- Next message: [Beowulf] file IO benchmark
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Eugen Leitl wrote: > On Thu, Nov 24, 2005 at 02:49:08PM +0100, Joachim Worringen wrote: > > >>>I've always liked Bonnie++ from fellow Melbournite Russel Coker: >>> >>> http://www.coker.com.au/bonnie++/ >> >>Well, the problem is that this is single-process, single-file benchmark, >>which is often very different from the I/O that a parallel application >>performs. > > > Is iozone better in that regards? I find its output more difficult > to parse, even if dumped as an Excel sheet and visualized. I/O is a complex topic, and a single number rarely tells you enough if you want to analyze what's going on. For this purpose, IOzone is very powerful. But again, it does only run on a single node (although can fork processes/threads to perform concurrent operations). You could start it on multiple nodes, but then the I/O is not synhronized, which may or may not what you want to do. For the purpose of the OP, IOzone is way to complex. Joachim -- Joachim Worringen - NEC C&C research lab St.Augustin fon +49-2241-9252.20 - fax .99 - http://www.ccrl-nece.de
- Previous message: [Beowulf] file IO benchmark
- Next message: [Beowulf] file IO benchmark
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Beowulf mailing list
