Cambridge - John Hill
Liverpool - John Bland, Stephen Jones, Mark Norman
RAL - Shaun de Witt, James Adams, Jens Jensen, Brian Davies
Oxford - Ewan Mac Mahon
Lancaster - Matt Doidge
RHUL - Govind Songara
Imperial - Duncan Rand


1. Puppet useful, object oriented - possibly useful to have a tutorial.
   
   Nagios for DPM - if it works along existing Nagios efforts.

   Monitoring in general?

   Network - architecture, tuning.  Maybe sites should record their
   settings.

   MySQL - tuning?  Five or ten tricks, everybody should know about the
   basics.
 
2. "Paper exercise" on evaluating technologies, interesting things from
   hepsysman, such as Lustre and CEPH.  Three candidates for physical
   evaluation with preprod service by March 2013.

   Aim is to split disk and tape storage into working and archive
   repositories respectively.  Lustre has the potential to improve on
   this.

   CASTOR is poor at draining disk servers, this can take 2-3 weeks.
   Lustre draining times?

   Currently in T1 internal wiki.  Will be moving either to the STFC
   public wiki or to the GridPP wiki.  Also other non-GridPP participants
   will be interested.

3. Deduplication?

   LOCKSS... are the more exotic methods of keeping data useful to
   investigating

   Worrying about the format/content of the file?

   Can we improve on support for access patterns?  While most of our
   configuration settings are "manual" eg buffer sizes, dCache hotdisk
   is an example of a system reacting to the access pattern and
   improving the access times.

4. AOB

   NOB


[09:59:55] John Hill joined
[10:01:11] James Adams yep, bit quiet
[10:03:47] Shaun de Witt joined
[10:04:27] James Adams left
[10:04:35] John Hill left
[10:04:38] Shaun de Witt No sound?
[10:04:40] Jens Jensen I lost sound..
[10:04:40] John Hill joined
[10:04:54] James Adams joined
[10:04:59] James Adams nope
[10:05:08] James Adams all gone silent for me
[10:05:10] John Bland yes
[10:05:11] James Adams yes, can hear jens
[10:05:13] Stephen Jones on and off
[10:05:15] James Adams and Shaun
[10:05:31] Mark Norman joined
[10:05:35] Ewan Mac Mahon joined
[10:05:35] Matthew Doidge joined
[10:05:57] Matthew Doidge left
[10:06:49] Ewan Mac Mahon Probably
[10:07:56] John Bland NFS4, maybe
[10:09:23] Ewan Mac Mahon We should certainly re-assess what settings are best
[10:09:35] Ewan Mac Mahon (also I seem not to have working outgong sound atm)
[10:10:16] Ewan Mac Mahon And remind people to check - I've just discovered that while the old gLite YAIM set them (badly) the EMI one doesn't do anything with them.
[10:10:48] Ewan Mac Mahon Yup. We should have a 'standard' DPM mysql config on the wiki or the like.
[10:11:04] Govind Songara joined
[10:11:11] Ewan Mac Mahon Nost of the benefit comes from a handful of basic options.
[10:11:42] Duncan Rand joined
[10:12:16] Ewan Mac Mahon brb - having EVO issues
[10:12:19] Ewan Mac Mahon left
[10:14:09] Brian Davies joined
[10:17:11] Duncan Rand it would be interesting to know about lustre draining times
[10:18:15] Duncan Rand yes, try it and see!
[10:21:21] Ewan Mac Mahon joined
[10:22:48] Ewan Mac Mahon We're mostly checking files that come into our storage anyway though aren't we?
[10:23:02] Ewan Mac Mahon Though I'm not sure you'd want to de-duplicate based on adler32
[10:23:48] James Adams sure not mutually exclusive, you want enough replicas for performance and safety but not more
[10:24:42] James Adams simple = good
[10:26:38] James Adams I recommend reading the presentation, it was quite comprehensive
[10:27:18] Duncan Rand EOS?
[10:27:43] Duncan Rand which presentation?
[10:28:27] Shaun de Witt Not EOS - CERN position is officially still not supported outside cern
[10:29:18] James Adams Beyond RAID presentation http://indico.cern.ch/getFile.py/access?contribId=38&sessionId=4&resId=0&materialId=slides&confId=160737
[10:29:54] Duncan Rand ok thanks
[10:31:39] James Adams noooo
[10:31:49] Duncan Rand work on ceph
[10:32:06] James Adams bye
[10:32:07] John Hill left
[10:32:08] John Bland bye
[10:32:08] Govind Songara left
[10:32:09] James Adams left
[10:32:09] Shaun de Witt left
[10:32:10] John Bland left
[10:32:12] Matthew Doidge left
[10:32:19] Brian Davies left
[10:32:22] Mark Norman left