[Bioclusters] Ethernet Performance

Tim Cutts bioclusters@bioinformatics.org
Tue, 13 Jul 2004 15:15:37 +0100

On 13 Jul 2004, at 8:51 am, Rene Storm wrote:

> Is there someone brave out there with pvfs2 performance results?

No, but Guy's talk from Bioclusters will tell you our experiences with  
IBM's GPFS.  Each chassis of 14 blades has its blastable data on a 600  
GB GPFS filesystem shared across all the machines.  Performance often  
actually exceeds that of the local disks that contribute to the  
filesystem (gigabit ethernet has better bandwidth than a local 2.5" IDE  
disk, after all)  Here's his talk:



Dr Tim Cutts
Informatics Systems Group, Wellcome Trust Sanger Institute
GPG: 1024D/E3134233 FE3D 6C73 BBD6 726A A3F5  860B 3CDD 3F56 E313 4233