Changes between Initial Version and Version 1 of GpuWorkFetch


Ignore:
Timestamp:
Dec 23, 2008, 11:24:31 AM (16 years ago)
Author:
davea
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • GpuWorkFetch

    v1 v1  
     1= Work fetch and GPUs =
     2
     3== Current policy ==
     4
     5 * Do weighted round-robin simulation, get per-project and overall CPU shortfalls
     6 * If overall shortfall, get work from project with highest LTD
     7 * Scheduler request includes just "work_req_seconds".
     8
     9Problems:
     10
     11There may be no CPU shortfall, but GPU is idle
     12
     13If GPU is idle, we should get work from a project that potentially has jobs for it.
     14
     15If the project has both CPU and GPU jobs, we may need to tell to send only GPU jobs.
     16
     17LTD isn't meaningful with GPUs
     18
     19== New policy ==
     20
     21Keep track of which projects can use which GPUs.
     22
     23Scheduler request includes