Present:
	Glasgow: Gareth, David
	Cambridge: John
	Oxford: Ewan
	RHUL: Govind
	Edinburgh: Wahid
	QMUL: Chris
	RAL: Jens, Brian


General round table on things which people are currently working on or
find interesting at the moment:

Gareth - no news

David - brand new disk servers slightly problematic, still on SL4 but
will be moving to SL5.

Ewan - upgraded head node to 1.8.3 - discovered that the workaround
described in the release notes was essentiall... caveat!

Brian - heavy ATLAS use again.  Working on optimising transfers.

Wahid - xrootQM, still testing WebDAV on DPM, it works... test
framework for dCache could be useful for DPM.

Chris - how much of a datafile is being used - maybe <10%, for
analysis.  Interesting how CERN are outsourcing their data centre, and
moving away from not-invented-here.

Note the federation stuff - will it be useful?  There is some new
monitoring stuff from the US.  Jeff Templon had done some work on
caching.  New xroot version needed for "federation".

Jens - interesting questions?  How does remote direct data placement
work in NFS4?  Discovered a file on home PC with NULs in it (Ewan had
a suggestion in chat).  Some IETF stuff on decoupling security from
data transfer might be interesting - idea is to speed up the latter by
first speeding up the former independently.  Also came across something
on "virtual data center"

Curation stuff - is HEP lagging behind?  Tend to use more odd
(proprietary) data types (like ROOT) rather than common ones (like
HDF, say).  There is a HEP curation WG.




[10:02:38] John Hill joined
[10:02:39] John Hill left
[10:02:52] John Hill joined
[10:04:23] David Crooks joined
[10:04:56] Gareth Roy Not really familiar with it
[10:05:21] Ewan Mac Mahon joined
[10:08:10] John Hill left
[10:08:13] John Hill joined
[10:08:18] Govind Songara joined
[10:08:37] Brian Davies joined
[10:11:22] Wahid Bhimji joined
[10:11:39] Wahid Bhimji sorry I'm late - I'm ill 
[10:13:30] Christopher Walker joined
[10:15:10] Ewan Mac Mahon There comes a point where the stuff is small enough to just give all the interested parties a copy on a USB stick.
[10:16:59] Wahid Bhimji did you see the data preservation chep talk 
[10:17:26] Ewan Mac Mahon Null bites == XFS crash
[10:17:31] Ewan Mac Mahon (often)
[10:17:33] Jens Jensen Haven't seent he chep presentations yet
[10:17:37] Jens Jensen not all of them
[10:17:40] Jens Jensen only a few
[10:18:12] Christopher Walker Roger also had a poster on data preservation
[10:18:16] Jens Jensen Could be XFS... I think I had the files on an XFS file syetm at some point.
[10:26:44] Ewan Mac Mahon Plus we're going to have the read-ahead size issue that we did ages ago with DPM;
[10:26:59] Ewan Mac Mahon where the read-ahead can mean reading loads more stuff than you need to.
[10:27:45] Wahid Bhimji yes - well we are supposed to use vector reading where you only read ahead stuff that you actually need to
[10:28:22] Wahid Bhimji that does work with DPM / rfio but didn't really seem to offer the required improvement to allow direct reading again so we still copy 
[10:28:43] Wahid Bhimji maybe with xrootd/DPM things will be better (with the new server) 
[10:30:13] Govind Songara left
[10:30:15] Gareth Roy left
[10:30:16] David Crooks left
[10:30:30] Wahid Bhimji left
[10:30:33] John Hill left