OGF Public Comments on StAR

Posted by: jcgordon 2012-06-17 07:52:47SITE field required by EGI and WLCG

The key site identifier in the EGI and WLCG e-infrastructures is the SiteName as defined by GOCDB. This should be added to the StAR definition.

While the storage element fqdn could possibly be resolved to a site name by software using the accounting data, the definition should be under the control of the site when publishing, not subject to later action by other parties.

Posted by: jcgordon 2012-06-17 08:04:53StorageResourceCapacityAllocated proprty to be added

In the StorageUsageBlock the property StorageResourceCapacityAllocated should added alongside StorageResourceCapacityUsed and StorageLogicalCapacityUsed.

By analogy with wallclock time signifying a cpu resource being blocked from other use and so a reasonable thing to be accounted and charged, storage space allocated is unavailable to other users on many systems.

It should be an optional property to be used by those who wish to account and charge for space allocated instead of, or as well as, space used.

An example use case would be a site who allocates whole disk servers or whole instances of storage systems to VOs, not sites who merely give VOs quotas in a bigger storage system.

This comment is on behalf of EGI and WLCG.

Posted by: jonkni 2012-06-21 03:01:34Comment originally posted by hartdavidl 2012-02-22

In addition to the fields listed, for accounting purposes, the record may want to permit the inclusion of a "charge" that is a function of the capacity used, in the same way that HPC job charges are often calculated as core-hours or node-hours modified by a queue or priority factor. In storage accounting, the bytes stored could be modified based on the StorageShare (e.g., slower disk vs. faster disk), StorageClass (e.g. pinned, replicated classes cost more), or MediaType (e.g., disk costs more than tape).

Second, the record does not permit accounting for storage "activity", i.e. the bytes and/or files read and/or written by respective users, if this is of interest (or available). If such values are supported, the forward-looking ValidDuration field may need to be redefined or replaced as a retrospective TimeDuration field, since the recorded reads/writes would apply in the TimeDuration leading up to the MeasureTime.

Posted by: jonkni 2012-06-21 04:24:20Reply to comments posted by jcgordon 2012-06-17

Hi John,

Thanks for input! The requirements for "Site" and "StorageResourceCapacityAllocated" fields are duly noted. While I'm a little reluctant to make the needed changes to this informational document, I propose that we make a new revision of StAR within EMI where these fields are included, and then make sure these requirements are covered in UR2.0.

-- JonKerrNilsen - 21-Jun-2012