Print

Print


> Dear colleagues,
> please let us know what other kinds of job execution requirements for the
> batch system should be expressible in the JDL and subsequently forwarded
> to the batch system.
>

First:
All requirements specified by the user or some other component of the LCG system 
should be passed as far as the jobmanager/BLAHPD interface scripts unless there 
is a really good reason not to.  This is so that site admins can customise
the scripts

Requirements:
RAM/Swap Required

Scratch disk space required

Speed of CPU required

CPU Features required (including CPU or virtual machine architecture).

Libraries required (shared and/or development with version number and ABI where appropriate).

Compilers required

Kernel version required

OS/Distribution/Variant (with version number)

Standards Compliant Environment required (POSIX,LSB,SUS or whatever with version and
whether the environment must be certified compliant or merely best efforts).

MPI interconnect speed requirements(as well as 1G Ethernet UCL has some nodes
with Myrinet and an Altix.  Neither of these has ever been used for LCG but the
potential is there).

Most of these should default to don't care but it might make sense
for CPU architecture to default to IA32 if the user didn't specifically
request otherwise.

In the other direction it would be good to know if a given job is known
to be checkpointable and if so what checkpointing systems are known to
work with it.

It would probably be good if these requirements could be flagged as
hard requirements or merely desirable as well. 

I would suggest that as many attributes as possible be translated
or you just end up with the lowest common denominator.  

William Hay


> ---------- Forwarded message ----------
> Date: Thu, 8 Sep 2005 11:16:41 +0200 (CEST)
> From: Massimo Sgaravatto - INFN Padova <[log in to unmask]>
> To: Maarten LITMAATH <[log in to unmask]>,
>      David Smith <[log in to unmask]>,
>      "Charles (Cal) Loomis" <[log in to unmask]>
> Cc: Francesco Prelz <[log in to unmask]>,
>      Giuseppe Fiorentino <[log in to unmask]>
> Subject: Forward requirements to local batch system
> 
> Dear all
> 
> I have just been given the task to try to better undestand the 
> requirement:
> 
> "Forward requirements to local batch system" 
> 
> set in the JRA1 workplan (looks like there were some misunderstandings 
> concerning who had to take care about that).
> 
> 
> As far as I am told, the issue was raised in particular by Cal. I think 
> that also David and Maarten are the right candidates to start the 
> discussion with.
> 
> So, could you please elaborate such requirement ?
> 
> 
> Do you mean that the various constraints expressed in the "JDL 
> requirement" expression should be translated in the proper batch system 
> directive and used when submitting the job, or were you thinking about 
> something different ?
> 
> 
> How about different batch systems supporting a different set of 
> attributes ?
> Should all the supported (by the used batch system) attributes be 
> translated, or only a minimum common (supported by all the supported 
> batch system) set of attributes should be considered ?
> 
> 
> How about translating requirement expressions specified on multiple CE 
> attributes, as the following example (taken from a real atlas production 
> job):
> 
> Requirements = other.GlueCEPolicyMaxCPUTime >= ( 14000000 / 60 ) / 
> other.GlueHostBenchmarkSI00 ) 
> 
> ?
> 
> Let's start the discussion :-)
> 
> 			Cheers, Massimo
>