[GE users] Re: pe's with SGE 6.0

Reuti reuti at staff.uni-marburg.de
Tue Mar 15 17:12:42 GMT 2005


Hi,

for PVM it seems, that the supplied aimk for building the programs in 
pvm/src are based to look for the PVM dir according to the SGE naming, 
but when you build PVM on your own, you get a different naming by 
default for these dirs in your pvm3/lib/ and pvm3/bin/.

I'll try to look into it in the next days, and adjust the things to get 
a more out-of-the-box behavior.

Cheers - Reuti


Chris Dagdigian wrote:
> 
> Hi Glen,
> 
> Parallel environments (PE's) are "mostly" the same in Grid Engine 6 vs 
> 5.3 in my experience.
> 
> The main "gotcha" difference is that in SGE 6 you tell the *qeueue* the 
> list of PE's it is able to support while in SGE 5 the opposite occured 
> -- the PE itself was configured with a list of queues that it was active 
> in. The other addition is the "urgency_slots" param (I think) which was 
> not in SGE 5.3.
> 
> If you had PE definitions or deployment scripts that worked in SGE 5.3 
> but not in 6 it may be due to the above. The "pe_list" parameter has 
> moved from the PE object itself and into the queue configuration.
> 
> For SGE 6 there are still the usual PVM and MPI templates and examples 
> that come with the distribution.  Just look in $SGE_ROOT/pvm/ and 
> $SGE_ROOT/mpi/.
> 
> Reuti also just updated the Grid Engine tight LAMMPI HOWTO which is here:
> 
> http://gridengine.sunsource.net/project/gridengine/howto/lam-integration/lam-integration.html 
> 
> 
> Back to PE's ...
> 
> This is what a generic loosely integrated MPICH PE would look like in 
> SGE 6:
> 
>> workgroupcluster:~ admin$ qconf -sp  mpich
>> pe_name           mpich
>> slots             512
>> user_lists        NONE
>> xuser_lists       NONE
>> start_proc_args   /common/sge/mpi/startmpi.sh $pe_hostfile
>> stop_proc_args    /common/sge/mpi/stopmpi.sh
>> allocation_rule   $fill_up
>> control_slaves    FALSE
>> job_is_first_task TRUE
>> urgency_slots     min
> 
> 
> Note that there is no list of queues that the PE runs in. This has moved.
> 
> The "pe_list" is now part of the queue configuration:
> 
>> workgroupcluster:~ admin$ qconf -sq all.q
>> qname                 all.q
>> hostlist              @allhosts
>> seq_no                0
>> load_thresholds       np_load_avg=1.75
>> suspend_thresholds    NONE
>> nsuspend              1
>> suspend_interval      00:05:00
>> priority              0
>> min_cpu_interval      00:05:00
>> processors            UNDEFINED
>> qtype                 BATCH INTERACTIVE
>> ckpt_list             NONE
>> pe_list               make mpich
>> rerun                 FALSE
> 
> < .... SNIP .... >
> 
> 
> I've tried to list the differences between Grid Engine 5 and Grid Engine 
> 6 at this URL:
> 
> http://bioteam.net/dag/gridengine-6-features.html
> 
> Not sure if I got it all but feedback/corrections are welcome.
> 
> 
> 
> 
> Regards,
> Chris
> 
> 
> 
> 
> 
> 
> Glen Otero wrote:
> 
>> I think I broke something while playing with grid engine 6.0, 
>> pvm-3.4.4-19, and mpich2. Anyone have pvm and mpi/mpich templates that 
>> they know work in creating pe's with SGE 6.0?
>>
>> Thanks!
>>
>> Glen
>>
>> Glen Otero Ph.D.
>>
> 


---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net




More information about the gridengine-users mailing list