Minutes of the storage EVO meeting 24 Nov 2010 Attending: Glasgow: David Edinburgh: Wahid Manchester: Alessandra Lancaster: Matt Bristol: Winnie Sheffield: Elena Liverpool: Stephen QMUL: Chris Sussex: Jeremy RAL T1: Brian, James, Jens (chair+mins) 0. Review of actions 1. Recommendations for sites - things we can do to improve chances of everything ticking along happily over Christmas break * Don't change anything! Although APEL has to be installed. * Maybe check free space - but ATLAS should be cleaning automatically. 2. Metrics discussion We have discussed this before, and now is the time to look at it again; how some metrics are currently more useful than others. Metrics are used by the PMB to see that we are making progress (or at least doing useful stuff). So we need to ensure that (a) they measure something useful and (b) they measure something we can influence. Brian points out the other numbers are useful too, like the transfer success rate. While most of those are outside our control, it is useful to be able to see how many transfers are failing etc. Jens will circulate proposal to the list. 3. Lustre Jeremy joined. Rebuilt Lustre client, with QLogic infiniband cards. Tests involved dd'ing a 2GB file from 120 cores which achieved a rate of 35 MB/s (ie 4200 MB/s). Oracle interested in selling ZFS+Lustre, possibly resulting in a fork, one proprietary Lustre and one open. QMUL upgraded yesterday. Sussex - intend to run StoRM, not deployed yet. Probably "safer" choice than BeStMan in the sense that we already have several sites in the UK running StoRM. Apropos StoRM, Chris reports that StoRM 1.6 is in beta, will work on SL5. What do we know about the European Lustre consortium? Not much, Jeremy will check his contacts. 4. AOB Some general discussions arising out of Jeremy's report about RAID controllers and the performance of Dell's H700 and H800. Jeremy has performance numbers with IOZone which he will circulate to list. Chris had a report that "software RAID is at least as effective if you have the bandwidth to the disks." Can this be corroborated or tested? 401 02/06/2010 Clean up the wiki ALL Low Open 416 10/11/2010 Report on Areca problems Sam Med Open Status unknown. 417 24/11/2010 Send GridPP25 input to Jens Sam High Open Done, closed. 418 24/11/2010 Check space token changes for 2010 with ATLAS Brian High Open Closed. Pre-allocation may change over the new year - Wahid had a mail from Kors about it. Will circulate relevant part of mail to list. NEW ACTIONS 419 08/12/2010 Circulate ATLAS mail on preallocation to list Wahid Med Open 420 08/12/2010 Circulate Lustre performance numbers to list Jeremy Med Open 421 08/12/2010 Circulate metric discussion to list Jens Med Open 422 08/12/2010 Check membership of European Lustre consortium Jeremy Med Open [09:59:07] Stephen Jones joined [09:59:43] Pete Gronbech joined [10:00:12] Winnie Lacesso Dry but suprisingly cold! [10:00:28] Brian Davies joined [10:00:32] Winnie Lacesso Yes, it's VERY pretty here too [10:01:13] David Crooks joined [10:02:43] Matthew Doidge joined [10:05:32] Jeremy Maris joined [10:06:53] Jeremy Maris hello my jkoala ava is working today [10:06:54] Alessandra Forti joined [10:08:42] Elena Korolkova joined [10:15:22] Queen Mary, U London London, U.K. joined [10:22:22] Queen Mary, U London London, U.K. Note Lustre 1.8.5 has just come out [10:24:02] Queen Mary, U London London, U.K. 500 000 dollars/year IIRC [10:28:56] Wahid Bhimji finally SL5 storm ! [10:37:07] Winnie Lacesso Pete & Jeremy, you're speaking very very very softly - please speak up! [10:39:50] Wahid Bhimji matt if you want to hang around after we can talk about your servers [10:40:04] Matthew Doidge will do [10:40:06] David Crooks left [10:40:07] Winnie Lacesso left [10:40:12] James Thorne left