[Beowulf] queue management systems survey

[Alexander Zubkov]

Hi all!
I making some type of analysis of queue management systems and will be 
glad to hear answers on survey, I prepared. I know that specifications 
and characteristics can be "extracted" from official manuals, but here 
I'm interested in practical cases. Thanks in advance.

1) Which queuing system You use at your cluser (OpenPBS, Torque, Sun 
Grid Engine, ...)? Mostly interest is queuing system, but scheduler will 
be interesting too.

2) Characteristics of your cluster:
- number of nodes
- type of nodes (workstations/dedicated, homogenous/heterogenous, cpus)

3) Queue characteristics:
- number of users
- average number of tasks in queue
- average number of running tasks
- only processor time taken into account or other resources too (memory, 
disk space, ...)

4) Common tasks characteristics:
- how much procesors required
- how much time required
- type of parallelism: exclusion of concurrent run of other tasks on 
assigned processors/something like SETI at home or United Devices projects

5) What is the pros and cons for this queuing system? Why it was chosen 
instead of others for your cluster?

PS. Some of this questions are related to scheduler too. But I need it 
mostly to recognize the capabilities of the queuing system. I.e. even if 
we have outstanding scheduler and the queuing system is bad - the result 
will be bad too.
----
Alexander Zubkov