[GE users] SGE 6.2u5 Resource Quota Set Question.

wlee_hess wlee at hess.com
Tue Feb 23 20:16:22 GMT 2010


To group,

I've got a question about Resource Quota Sets.  I'm currently running two SGE servers with versions 6.1u3 (production setup) and 6.2u5 (Test and soon to be new production).

We currently have three separate files with each defining a resource quota set.  Does it matter if all resource quota set entries be in one file or in multiple files?  How does one or the other affect performance of the SGE qmaster/scheduler?  We're running roughly a 1500+ node Linux cluster and I was just wondering about this.  Currently, we have the following three resource quota sets in three separate files,

File: gnodes_for_testing
------------------------
{
   name         gnodes_for_testing
   description  "restrict use of nodes in hostgroup @test_gnodes to developers"
   enabled      TRUE
   limit        users {user1,user2} hosts {@test_gnodes} to slots=8
   limit        users {*} hosts {@test_gnodes} to slots=0
}

File: remote_gpu
----------------
{
   name         remote_gpu
   description  "Control access to Remote  gpu cluster"
   enabled      TRUE
   limit        users {@REMOTE_PROCESSING} hosts {remp*} to slots=8
   limit        users {*} hosts {remp*} to slots=0
}

File: max_per_host
------------------
{
   name         max_per_host
   description  "limit concurrent jobs to 8 on gpu systems and 2 on all others."
   enabled      TRUE
   limit        users {user4} hosts {n0135*,n1360.ihess.com} to slots=1
   limit        users {user7} hosts {n0145*,n01460.ihess.com} to slots=0
   limit        users {*} hosts {g000*} to slots=6
   limit        users {*} hosts {g001*} to slots=6
   limit        users {*} hosts {g*} to slots=8
   limit        users {*} hosts {*} to slots=2
}

If I take the above and reduce it to the following:

File: all_rule_sets
-------------------
{
   name         gnodes_for_testing
   description  "restrict use of nodes in hostgroup @test_gnodes to developers"
   enabled      TRUE
   limit        users {user1,user2} hosts {@test_gnodes} to slots=8
   limit        users {*} hosts {@test_gnodes} to slots=0
}


{
   name         remote_gpu
   description  "Control access to Remote  gpu cluster"
   enabled      TRUE
   limit        users {@REMOTE_PROCESSING} hosts {remp*} to slots=8
   limit        users {*} hosts {remp*} to slots=0
}

{
   name         max_per_host
   description  "limit concurrent jobs to 8 on gpu systems and 2 on all others."
   enabled      TRUE
   limit        users {user4} hosts {n0135*,n1360.ihess.com} to slots=1
   limit        users {user7} hosts {n0145*,n01460.ihess.com} to slots=0
   limit        users {*} hosts {g000*} to slots=6
   limit        users {*} hosts {g001*} to slots=6
   limit        users {*} hosts {g*} to slots=8
   limit        users {*} hosts {*} to slots=2
}

Does it make much of a difference?

Kind Regards,

Wayne
email:wlee at hess.com

------------------------------------------------------
http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=245710

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



More information about the gridengine-users mailing list