[GE users] limitation of the max slots or nodes

ron ron_chen_123 at yahoo.com
Fri Dec 5 07:05:59 GMT 2008


--- On Mon, 11/17/08, holy <holy8086 at gmail.com> wrote:
> I will use SEGE6.2 in large-scale grid.

As Rayson has already mentioned, SGE runs on systems with 4000 nodes (and that system almost has 64000 AMD cpu cores).

While there are many TOP500 supercomputers that have lots of nodes, many of them group nodes into partitions to form an hierarchy. Thus to the batch system, there are only a small number of "nodes" to control as the execution daemons run only on the partitions.

Regarding Ranger, as far as I know, there are 4000 execds running in one single SGE cluster, so I would say that SGE scales good compare to those hierarchically organizied clusters.


> When GE starts, some warning messages about file descriptor
> are found to the
> spool/qmaster/messages file.
> Is there some limitation of the maximum number of slots or
> number of nodes
> to be able to operate GE?

If you use Linux as your qmaster machine, you need to get a special build of SGE if you have over 8000 execution nodes.

With the next update of SGE 6.2, you should be able to get over that limit as we are enabling an enhancement by default. For earlier versions, you can compile with a special flag called "-DUSE_POLL" if you use the source distribution.

 -Ron




> When the number of nodes or slots in a CELL are increased,
> would the number
> of file descriptor(and so on) be consumed?
> 
> Thanks,
> holy
> 
> ------------------------------------------------------
> http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=88883
> 
> To unsubscribe from this discussion, e-mail:
> [users-unsubscribe at gridengine.sunsource.net].

------------------------------------------------------
http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=91282

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



More information about the gridengine-users mailing list