Search results: Spectrum LSF

Showing 149 of 2943

Enhance GSLA to enable in-queue preemption

The Guaranteed SLA system allows sharing of Resources defined by ServiceClasses. We would like to use that to allow two or more groups to pool and share compute slots, wherein they typically unused slots would be available to all of the groups. Th...

almost 5 years ago in Spectrum LSF / Scheduling 1 Not under consideration

6 VOTE

Multiple esubs (master + remote) in a multicluster setup

We use LSF in a multicluster configuration. We also use an esub to dynamically assign which jobs are sent to each cluster. As far as I'm aware, only one esub (on the master) can be used. We would like to have functionality for esubs on the remote ...

almost 10 years ago in Spectrum LSF 2 Not under consideration

4 VOTE

Need to submit jobs from x86_64 systems to CORAL systems

The IBM supercomputer is only one part of the computing ecosystem in the center. We need the ability to submit jobs back into the cluster from other platforms in the center that are x86_64.

almost 7 years ago in Spectrum LSF 0 Delivered

4 VOTE

Adjustable Number of MPS Daemons for GPU Sharing

For now, IBM Spectrum LSF allows automatically instantiating one single MPS Daemon when requested through the '-gpu' directive:---gpu 'mps=yes:[...]'-- However, it has proved beneficial in several performance benchmarks conducted on different clie...

about 7 years ago in Spectrum LSF 1 Delivered

1 VOTE

Add support for GPU jobs in brun

Currently, the brun command can force a non-gpu job to run using various numbers of CPU's on one to many hosts, but this is not possible for GPU jobs. Add the ability to specify the GPU's or gpu_id's for a forced run gpu job to use, and follow the...

almost 2 years ago in Spectrum LSF / Administration 1 Not under consideration

3 VOTE

License Scheduler fix INUSE license token detection on same-server Jobs

The License Scheduler is unable to correctly identify the INUSE and RESERVE values of requested tokens (e.g. rusage[MYTOKEN=4], when more than one jobs run on the same server by the same user. Possible solution:This could be solved, if the License...

over 5 years ago in Spectrum LSF 2 Not under consideration

1 VOTE

Add a 'bsub -Gl limit_name' option to separate Fairshare Assignment from the Limit also add a default limit pattern to use if ONE_UG_TREE does not find a matching limit

We have many cases, where we have limits for users, but we also want to want to apply a fairshare group for the job to arrive at a base allocation. Unfortunately, with ONE_UG enabled, if the Fairshare Group does not have a corresponding Limit, not...

almost 2 years ago in Spectrum LSF / Administration 0 Future consideration

1 VOTE

be able to add (and remove) new type of metadata

to manage other metadata that the ones already defined: LSF user extra limit & TTL, license quota, etc ... currently, we must update several scripts & conf files to add (or remove) other metadata type than the built-in ones.

almost 2 years ago in Spectrum LSF / Operations/RTM 0 Planned for future release

5 VOTE

reject multiple bmod/bkill's on same array within a time frame

In reference to PMR 59023,019,866 , our LSF cluster has frequently halted and the scheduler spirals down due to a user running too many bmod/bkill on large job arrays. Despite internal training sessions and company wide communications, this issue ...

almost 10 years ago in Spectrum LSF 5 Delivered

2 VOTE

When a Benchmark job fails It would be nice to see the bsub used to submit the job

We can manually construct the actual bsub, but it would be nice to be able to cut and paste it when editing the Benchmark job so we can test that it actually will work. In one case we were getting "Queue Permission Errors" and it having that bsub ...

about 4 years ago in Spectrum LSF / Reporting/Explorer 2 Not under consideration

Please enter your email address

FILTER BY CATEGORY