Dear Fotis,

Sorry for the late answer but your email got lost in the wrong folder...
On line some answer.


On 05/10/2013 04:37 AM, Fotis Georgatos wrote:

Dear Usage Record working group,

First, I would like to ask you, according to your experience, which
known technologies are closer to implementing the described v2.0 Usage Record;
(the question regards both cloud and scheduling systems and an answer
would be nice if it is of the type "we are using X with open source lib Y,
yields 80% of the standard, while we do billing based on A & B procedures)
ie. if someone holds a working prototype knowing about it would be great.


The work done for UR2.0 take in account work done separately in other projects especially within EMI where other, more specific standards were defined (CAR for computing and StAR for storage). There is also another definition from EGI (fedcloud task force for cloud accounting). As you can see from the Attribute Matrix at the end of the UR2.0 document, the attributes of the other definition have a correspondence in this document.
At the moment I do not know any implementation of UR2.0 but it should be quite straightforward to support it in products that support one of the other definitions mentioned. The other definition are indeed already being supported in several different products (for example CAR for DGAS and APEL, StAR and the EGI cloud definition in the new sensors that are in development for the storage and cloud).



As regards the standard itself 2 comments follow, for potential OPTIONAL elements:
* It would be good to provide a field somewhere for the so-called "xfactor";
 it is being used across a few HPC sites as a post-processing metric for jobs:
 http://docs.adaptivecomputing.com/maui/5.1.2priorityfactors.php#xfactorsub
 Typically its value is in the 1..2 range and centers track the histogram closely.

I had a look at the link and I do not understand exactly why you would need this value after the job is finished. And what is its purpose in the accounting?
Maybe it could be useful for future submission of job of the same kind but this looks like more monitoring to me.


* You may wish to provide the ability to report "container job ID", ref.
 http://oar.imag.fr/sources/2.5/docs/documentation/OAR-DOCUMENTATION-USER/#container-jobs
 As name suggests it permits to first make a reservation there run your jobs in it;
 it can be a useful feature for billing, eg. if you organize an HPC school 
 and, would rather wish to avoid charging twice, for trainee jobs within trainer's job.
 (just an example, they have many more uses)

I am not sure why this would be necessary. If the accounting takes is based only on the CPU time then there should be no problem on how many processes are or jobs are run. But maybe I am missing something here.


kind regards,
Fotis

ps. I am not receiving the m-list traffic, kindly keep me in cc for feedback.

In the last weeks there were not many emails on the list but, if you wish, you can register on the OGF web site:

http://redmine.ogf.org/

After registration just click on "Projects" on the top left and look for "Standards" -> "Management Area" -> "UR WG".

In the page of the group you can join by clicking on the right side of the page.


Kind regards,
Andrea




--
Drs./Eng. Fotis Georgatos <Fotis.Georgatos@uni.lu>
HPC Systems Engineer, LCSB (University of Luxembourg)
L-4362 Esch-sur-Alzette, Campus Belval, avenue des Hauts-Fourneaux 7
Phone:  +352 466644 5609






-- 
Andrea Cristofori
INFN-CNAF
Viale Berti Pichat 6/2
40127 Bologna
Italy
Tel. : +39-051-6092920
Skype: andrea-cnaf