Attending: Elena, Emyr, John H, Jeremy, Mark, Alessandra, Chris W, Chris B, Dave, Govind, John B, Matt, Rob, Robert, Steve, Wahid, Pete, Sam, Jens (chair+mins)

1. What use if any should we make of the proposed data storage workshop at EGI CF in Manchester in April?

We cannot use it as a DPM workshop or dCache workshop, because there will
now be both DPM and dCache workshops at ASGC/ISGC.  There is also IoP in
Liverpool at the same time.  The best opportunity seems to be to engage
with non-LHC data communities, either the users or the equivalent of us
in those communities (ie storage/infrastructure operators/supporters)

Alternatively, we organise a dedicated workshop covering "big science
data", later in the year, maybe focusing on UK-based storage
infrastructure or projects.

There are advantages and disadvantages of each - for the former, people
will need to register, but we also benefit from people who've just come
for the CF; for the latter, people are less likely to travel from outside
the country and probably wouldn't travel just to hear about GridPP -
there'd have to be enough content and contributions from non-LHC to make
it worth their while.  It should, however, be possible to obtain funding
for such an activity (EPSRC, STFC?)

Alternatively alternatively, we could organise a day at GridPP, but that'd
be more for our own benefit - others may be less likely to just travel for
a day??  But it would be advantageous because we would already have the
rooms.

It was agreed to try to contact SKA and Sanger - and other institutes
or communities we can think of? - to see whether they are interested
in collaborating with us on one of the two options.  There is also
HPC-SIG - would these be the Usual Suspects?

If there is not enough interest, then we may just have to drop the
workshop.  OTOH, given that GridPP is a co-organiser, it'd be good
to fly the GridPP flag a bit - but Neasan says we have a booth anyway
(albeit shared with the NES).

Jens will make the initial contacts to SKA and Sanger.


As regards the redirector discussion last week, we probably have enough
experience and volunteers to participate at the current level of xroot
redirection.  CMS will clarify their position tomorrow.  It would be
good to start collecting results.  Wahid will seek volunteers offline
for WebDAV redirection.

There will be a discussion at HEPiX in Bologna.  Using WebDAV instead of
xroot would have the (theoretic) advantage of being standards-based,
although we have seen interoperation problems before - this has been
discussed in previous meetings.  See also references in chatlog.


Regarding the agenda for GridPP, let's have a more extended discussion
next week.  It would be good to cover
* Future directions in storage/data mgmt and data access
* Networks
* Technologies
* Different aspects; like access patterns
* Use of cloud storage (or combined with cloud compute)
  - need some more testing/experience
* Big (science) data
- with "speculative but grounded" contributions


[10:02:02] Wahid Bhimji jens we can't hear you if you are trying to speak ...
[10:02:06] Jens Jensen Mic problems...!
[10:02:09] Jeremy Coles Jens - no sound from you.
[10:02:19] Wahid Bhimji ah ok - we'll standby for a minute
[10:02:34] Jens Jensen camera works
[10:03:03] Elena Korolkova yes
[10:08:15] Jeremy Coles PM: "perhaps this would be an opportunity to engage with "new communities" (i.e., non-LHC). For example, people who are now finding themselves with large amounts of data --- much more than they have every had to handle before --- could discuss what are their concerns and what they want to achieve. "
[10:09:29] Wahid Bhimji but the new communities - are still those within EGI
[10:09:52] Ewan Mac Mahon That's clearly right; we need to be getting hold of the Bio folks and the Astro folks and the like and tempting them with our stuff
[10:10:13] Ewan Mac Mahon But the EGI meetings may not be the best venues for that.
[10:10:41] Jens Jensen But at least we can talk to the ones who are there
[10:10:47] Wahid Bhimji yeah but I would rather do that with Industry
[10:10:59] Wahid Bhimji than EGI
[10:11:13] Ewan Mac Mahon Storage is one of our big advantages - CPU is easy, everyone can do that.
[10:11:33] Ewan Mac Mahon What we have that lots of folks don't, is a) data management, and b) federated authentication.
[10:12:28] Ewan Mac Mahon Yes. Yes it is.
[10:12:33] Jeremy Coles Agreed
[10:12:55] Ewan Mac Mahon Someone should actually go to Sanger and basically do the doorstep salesman routine.
[10:13:13] Ewan Mac Mahon They need to know what we've got, even if they don't know they need to know.
[10:13:20] Ewan Mac Mahon And they really do.
[10:13:27] Jens Jensen I talk to Sanger every now and then. They're running irods  
[10:14:44] Wahid Bhimji Actually - I probably wouldn't go if there wasn't the workshop - which is further why I am not massively keen
[10:15:17] Ewan Mac Mahon Or just do it at a later occasion.
[10:15:30] Ewan Mac Mahon e.g. the next or next+1 gridpp mettings.
[10:15:47] Ewan Mac Mahon Or just run a separate little event for potential new people.
[10:15:58] Ewan Mac Mahon It doesn't have to be tied in to the EGI forum.
[10:19:37] Ewan Mac Mahon Good grief, no.
[10:21:00] Jens Jensen ELIXIR...
[10:22:35] Wahid Bhimji this seems very different to the EGI w/shp idea... though the advatnage is that at least some storage system developers will be there
[10:22:48] Wahid Bhimji (though we can invite them back again for this super workshop)
[10:22:55] Jeremy Coles https://indico.egi.eu/indico/conferenceTimeTable.py?confId=679#all
[10:23:04] Ewan Mac Mahon The downside is that the someone else is also charging for the event.
[10:23:09] Jens Jensen Yes, poeple may already be travelling to the
[10:23:12] Jens Jensen event
[10:23:16] Ewan Mac Mahon Which is why I'm not going.
[10:23:16] Wahid Bhimji except for the organisation time
[10:24:42] Ewan Mac Mahon If people just register for the workshop then we're not getting much benefit from co-locating it.
[10:25:07] Alessandra Forti indeed
[10:25:13] Ewan Mac Mahon If anything, it'll be worse because the EGI meeting will be using lots of rooms.
[10:25:24] Ewan Mac Mahon And accomodation, and queing for coffee.
[10:25:26] Alessandra Forti the point of a conference is that you meet people who are already going
[10:25:27] Ewan Mac Mahon And stuff.
[10:25:47] Wahid Bhimji As I say they are going to ask us to a DPM workshop later in the year. We can make it an (ex)EMI workshop and colocate these other guys coming which everyone would be interested in
[10:26:45] Ewan Mac Mahon I think it's useful to separate our own 'internal' discussions of DPM/dmlite/srm-less SEs etc. from a more new community 'sales pitch' sort of thing.
[10:27:12] Wahid Bhimji ok - its just that they are the experts on our architecture
[10:27:17] Ewan Mac Mahon Partly because they're just different topics, and partly because I want to convince people that our stuff is good, and seeing the guts of it isn't necessarily the best way to do that.
[10:28:37] Sam Skipsey I should point out that I *am* going, and I *am* giving a talk, aimed at small VOs about how to use our storage.
[10:28:47] Ewan Mac Mahon Oh. Well that's that then.
[10:30:00] Ewan Mac Mahon Can you trail the idea of a possible future intro meeting and see if anyone sounds wildly enthusiastic about it?
[10:30:05] Sam Skipsey Of course.
[10:33:26] Wahid Bhimji I haven't been but thats my impression too
[10:33:40] Wahid Bhimji I think we should do it later and do it properly
[10:33:51] Jeremy Coles The discussion seems to have the same feel as that between Wahid and myself. The conclusion is that we should try but there is concern about wider participation.
[10:34:12] Ewan Mac Mahon I think so; we're not going to be able to pull a good one of these together for the EGI CF.
[10:34:33] Ewan Mac Mahon (that is, I agree with Wahid about doing it later and well)
[10:34:47] Christopher Walker https://www.gridpp.ac.uk/wiki/WebDAV - please keep it up to date
[10:35:06] Christopher Walker Also contains xrootd details
[10:35:15] Alessandra Forti Doing it later completely looses the advantage of a conference where people go also for other reasons
[10:36:34] Alessandra Forti I also have to comment on the fact that GridPP pushed to have this conference when NGS lost the funding and I find strange to have so little presence
[10:37:23] Jens Jensen https://indico.cern.ch/getFile.py/access?contribId=0&resId=0&materialId=slides&confId=218328
[10:38:11] Alessandra Forti manchester is setup
[10:38:23] Alessandra Forti I've updated that already
[10:38:32] Jens Jensen https://www.gridpp.ac.uk/wiki/WebDAV
[10:39:53] Wahid Bhimji finally loaded that talk - Fabrizio gave a talk at the last GDB
[10:43:11] Jens Jensen Any idea how much time we'll have for a storage/datamanagement slot?
[10:44:43] Wahid Bhimji technology strategy shouldn't restrict itself to storage
[10:48:00] Ewan Mac Mahon I still think we can make a perfectly good arguement that our storage IS cloud.
[10:48:23] Ewan Mac Mahon It does very much what 'cloud' systems like Amazon do.
[10:49:03] Ewan Mac Mahon Unless you think 'cloud' means 'has the EC2/S3 API'
[10:49:04] Alessandra Forti Grid is a IaaS cloud
[10:49:14] Alessandra Forti Infrastructure as a Service
[10:49:16] Sam Skipsey It should also be pointed out that "Cloud Storage" is "Storage", not "Data Transfer"
[10:49:41] Sam Skipsey I have yet to see any evidence that anyone actually solved the hard problems that our solutions approach.
[10:49:52] Alessandra Forti And there is also STaaS
[10:50:00] Alessandra Forti Storage as a Service
[10:50:19] Alessandra Forti bye