SourceForge : Post

Project: Editor Discussion > INFO:HPC Job Scheduling Use Cases > "Held" job state in "HPC Job Scheduling: Base Case and Common Cases" 2006-07-01 draft > List of Posts

Forum Topic - "Held" job state in "HPC Job Scheduling: Base Case and Common Cases" 2006-07-01 draft: (2 Items)

View: as

Update

Troy Baer

08/16/2006 10:30 AM

post5642

"Held" job state in "HPC Job Scheduling: Base Case and Common Cases" 2006-07-01 draft

Section 2.2:  State diagram -- This discussion neglects the "held" job state -- that is, a state where the job cannot 
run or is not considered to scheduling due to administrative/policy reasons.  This is not the same as "queued", in that 
a "queued" job may be considered for scheduling whereas a "held" job is not.  This is touched on tangentially in the 
discussion of job preemption/suspension in Section 3.8, but I would argue that "held" jobs should be considered as part 
of the base case.

Hiro Kishimoto

03/13/2007 4:37 PM

post5766

Re: "Held" job state in "HPC Job Scheduling: Base Case and Common Cases" 2006-07-01 draft

We discussed this comment at length in one of the calls at the end of last year, and we thought it was a reasonable 
comment, but that the group felt that it did not belong in the "base Case". We believe that our discussions with the 
implementers for the Supercomputing demo reinforced our belief that it was not appropriate to have it in the base case. 
Note that this is explicitly dealt with in the "Extensions", so we acknowledge this issue (we just believe that it does 
not belong in the base case).

So, in short, we respect this comment, and we appreciate the person taking the time to make this comment, but we don't 
believe that a modification to the document is warranted. So we would request that the document as written be published 
as an OGF Informational Document.

Marty Humphrey, OGSA-HPCP-WG co-chair

Return