[GE users] Reversing Sequence Number Sorting?

lvta0909 lvta0909 at gmail.com
Fri Aug 20 01:32:07 BST 2010


On Thu, Aug 19, 2010 at 7:26 PM, reuti <reuti at staff.uni-marburg.de> wrote:
> Hi,
> Am 19.08.2010 um 23:44 schrieb jlforrest:
>> I have a queue that I've configured to
>> sort jobs by queue instance sequence
>> number. This works fine.
>> However, one of the people I work with
>> observed that this setting results in the
>> nodes with lower sequence numbers
>> running more jobs than the higher sequence
>> number nodes when the queue isn't
>> completely full. He was wondering if there
>> might be long term physical damage if some
>> nodes, e.g. the low sequence number nodes,
>> are always 100% busy. This probably generates
>> way more heat.
> yes - correct. You could even start to switch off the nodes with the higher numbers by SDM/Hedeby automatically.
> What I heard, is that thermal expansion and shrinking is most stress to the material. So, when some nodes are always running at 100% it's best, as they won't change temperature (and they should be designed to run under such conditions). And the nodes at the other end are always colder - also fine.
>> Is there a way to sort by descending sequence number?
>> Changing this parameter every year or so might
>> be a good thing to level out wear and tear.
> When you have a maintenance contract for the hardware, it wouldn't matter. You could also argue, that you have some less used nodes left at the end of the maintenance.
> -- Reuti
> PS: It would nevertheless be possible to generate random sequence numbers each night - but as `qstat` would also change the order, it can get confusing. I'm not aware of any option to just change the order to descending.

[not sure if its works, but ] another idea is to change the
sort_method to "load",  and change the load_formula to seq_no, or
-seq_no. You could, as Reuti suggested, change this every night.
Another ideia is to define a new "complex resource" that computes a
number for each host by the date.

sorry if I'm suggesting nonsense things here, I'm just a newbiew in GE, heh.

>> Cordially
>> --
>> Jon Forrest
>> Research Computing Support
>> College of Chemistry
>> 173 Tan Hall
>> University of California Berkeley
>> Berkeley, CA
>> 94720-1460
>> 510-643-1032
>> jlforrest at berkeley.edu
>> ------------------------------------------------------
>> http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=275556
>> To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].
> ------------------------------------------------------
> http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=275567
> To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].


Tiago Bitarelli Gomes
lvta0909 at gmail.com


To unsubscribe from this discussion, e-mail: [users-unsubscribe at gridengine.sunsource.net].

More information about the gridengine-users mailing list