[jsdl-wg] IndividualCPUCount
Peter G. Lane
lane at mcs.anl.gov
Tue Aug 22 09:55:28 CDT 2006
Andreas Savva wrote:
> Could you put this on the errata tracker so we don't forget?
Done.
> I think you are reading a lot more into the difference in wording than
> was intended.
Perhaps, but I don't know ahead of time what the intent of the authors is. ;-)
Peter
> Andreas
>
> Peter G. Lane wrote:
>> Would it be possible to change the description of IndividualCPUCount to
>> say simply that it specifies the required number of CPUs instead of the
>> number to be allocated. The way it's worded now, match making based on
>> the actual number of CPUs (i.e Condor) seems improperly represented by
>> this element. Making it a little more general allows for the
>> interpretation that a host must at very least have a certain number of
>> CPUs, but it does not dictate the number of CPUs actually being
>> allocated for the job. Resource managers such as LSF and PBS that only
>> operate using CPU allocations could still be considered included in the
>> description because if an allocation is successful then obviously the
>> host must have that many CPUs.
>>
>> In addition, the following elements are worded similarly to
>> IndividualCPUCount so as to effectively prohibit purely match making as
>> opposed to strict allocation of host resources:
>>
>> IndividualVirtualMemory
>> IndividualDiskSpace
>> TotalDiskSpace
>>
>> For what it's worth, the rest of Resources' elements seem to be worded
>> more generally (i.e. "required by the job" as opposed to "allocated to
>> the job"). Derive from that what you will. ;-)
>>
>> Peter
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/x-pkcs7-signature
Size: 3804 bytes
Desc: S/MIME Cryptographic Signature
Url : http://www.ogf.org/pipermail/jsdl-wg/attachments/20060822/06263747/attachment.bin
More information about the jsdl-wg
mailing list