Giganet vs. Myrinet vs. gigabit

Greg Lindahl lindahl at conservativecomputer.com
Thu Oct 18 11:45:41 PDT 2001


On Thu, Oct 18, 2001 at 12:28:48PM -0500, Jarrod Smith wrote:

> In a more
> real-world situation, the Myrinet would be idling most of the time waiting
> on disk I/O unless perhaps you had striped RAID on both ends, or lots
> (10+) clients talking simultaneously to one very-high-performance disk
> array on the server.

Actually, in computational biology, one common I/O operation in gene
sequence matching is to have N clients read the same large input
file. For that, the network is the bottleneck.

But in general you're right, Myrinet is not the weakest link in the
chain, the disk on the server is.

greg




More information about the Beowulf mailing list