Hi Winnie,
For our latest 8-core nodes we have channel-bonded
the two 1GbE interfaces and used 1 network switch for 16 nodes
so that the bandwidth available per core is approximately the same
as our older (and slower) 4-core nodes which use just one GbE
interface and 40 nodes/switch. However, it should be noted that
the interswitch connections are only 10GbE and the total bandwidth
from the storage of ~50-60 Gb/s is a bit less than that to the compute
nodes ( 80 Gb/s ) so with a full uniform load the node-to-switch connections
won't be the bottleneck.
cheers,
Alex
On Friday 29 August 2008, Winnie Lacesso wrote:
> Question for those of you with 8-core WN, where something like
> gpfs is *not* in the equation - do you have these on a 1Gbps
> network or a 100Mbps network?
>
> Because one wonders, if WN are on a 100Mbps network could there be
> contention between several high-I/O jobs on that WN for that single
> 100Mbps eth0 NIC?
>
> Anyone seen this?
--
------------------------------------------------------------------------------
| |
| Dr. Alex Martin |
| e-Mail: [log in to unmask] Queen Mary, University of London, |
| Phone : +44-(0)20-7882-5033 Mile End Road, |
| Fax : +44-(0)20-8981-9465 London, UK E1 4NS |
| |
------------------------------------------------------------------------------
|