[Aces-support] ao jobs quit immediately

William Boos billboos at MIT.EDU
Mon Apr 2 15:31:15 EDT 2007


hi greg,
i meant the ds-nodes, as in /net/ds-02/scratch-1.  how can we find which
ds-nodes are local to a particular aces room?

-bill
 

-----Original Message-----
From: aces-support-bounces at mitgcm.org
[mailto:aces-support-bounces at mitgcm.org] On Behalf Of
aces-admin at techsquare.com
Sent: Monday, April 02, 2007 3:09 PM
To: ACES-support at mitgcm.org
Subject: Re: [Aces-support] ao jobs quit immediately

hello billboos-

the ao-nodes do have more ram than the
geo-nodes, which have faster busses (i think i remember that correctly). 

as far as scratch-disks are concerned, kinda means what you mean. 

each compute-node has a local /scratch
partition. this is the leftover of every root drive.

if you mean the ds-nodes, then that's an order of magnitude different. true,
they are all distinct machines, but they are located in the several
aces-rooms across campus. locality, then, may affect performance.

which do you mean ? 

[greg]


> From: "William Boos" <billboos at mit.edu>
> Date: Mon, 2 Apr 2007 13:40:48 -0400
> MIME-Version: 1.0
> Reply-To: ACES-support at mitgcm.org
> 
> When I submit a job to ao it starts and then quits immediately, for 
> both interactive and scripted jobs.  Examples are 58728.ao (a 1-node 
> interactive
> job) and 58729.ao (a 24-node job).
> 
> Also, I've been trying to run on ao because my jobs typically run in 
> half the time there as on geo.  Anyone know why this would be?  Are 
> the scratch disks mounted local to a particular cluster, thereby 
> making disk writes quicker for some disks?
> 
> -Bill
> 
> _______________________________________________
> Aces-support mailing list
> Aces-support at acesgrid.org
> http://acesgrid.org/mailman/listinfo/aces-support
> 
_______________________________________________
Aces-support mailing list
Aces-support at acesgrid.org
http://acesgrid.org/mailman/listinfo/aces-support




More information about the Aces-support mailing list