No subject


Wed Jan 12 20:38:46 GMT 2011


decided to use Share-Based Policy configuration (i.e. share tree) as
it support feedback scheme.

I'm not able to understand decay factor, half life factor,
compensation factor and their affect with share tree policy.

To test share tree policy, created two projects with 2 SGE user groups
(ACL). The sharetree is as follows:

# qconf -sstree
id=0
name=Root
type=0
shares=100
childnodes=1,2
id=1
name=adm_p
type=1
shares=70
childnodes=NONE
id=2
name=users_p
type=1
shares=40
childnodes=NONE

===============================

# qconf -sprj adm_p
name adm_p
oticket 0
fshare 70
acl adm
xacl NONE
# qconf -sprj users_p
name users_p
oticket 0
fshare 20
acl users
xacl NONE
# qconf -su adm
name    adm
type    ACL
fshare  0
oticket 0
entries root
# qconf -su users
name    users
type    ACL
fshare  0
oticket 0
entries user1,user2
#

I expected from this config, The jobs running under adm_p (70) should
use 70% of the CPUs or nodes of the cluster and the jobs of  users_p
should only use 20% of the cluster resources. But its not happening.
All the jobs will get running with FIFO order.

I think the the policy config is not complete or I might have
misunderstood the concept. Can some on the list throw some light to
understand on policy configuration.

Thanks in adavnce

------------------------------------------------------
http://gridengine.sunsource.net/ds/viewMessage.do?dsForumId=38&dsMessageId=154541

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



More information about the gridengine-users mailing list