Attending: Brian, Chris B, Chris W, David, Duncan, Gang, Gareth, Jens (c+m),
	John B, John H, Raja, Raul, Robert, Steve, Matt D, Elena, Alessandra,
	Ewan.

Apologies:
	Tom, Sam, Wahid, Jeremy


0. Operational issues - review

   No issues.

1. DPM workshop - https://indico.cern.ch/event/265334/

   As mentioned, the workshop is ongoing.  Wahid had suggested we get involved
   in HTTP.  We should have a report next week?

2. perfSonar:

This meeting may have a few interesting things to look at:
http://indico.cern.ch/event/289680/
[1] perfSONAR -
http://indico.cern.ch/event/289680/contribution/5/material/slides/

There's also a presentation from our very own Dave Kelsey but probably
doesn't contain anything new for hardened GridPPers.

3. V. quick review /reminder of web docs

4. Did anyone have much hands-on experience with SDN?

5. I still think it would be useful to do some stuff on how GridPP could
be said to do, or can be applied to, "big data" but was stymied a bit by
the lack of volunteers last week...

6. AOB
Steve Jones: (30/04/2014 10:03:48)
Is anyone saying anything, since Brian spoke?
Gang Qin: (10:04 AM)
I can't hear anything 
Steve Jones: (10:04 AM)
No. Nor me...
Duncan Rand: (10:07 AM)
what's atlas' median file size?
Raja Nandakumar: (10:07 AM)
10MB is what Brian said
Duncan Rand: (10:07 AM)
all the log files I think
Raja Nandakumar: (10:08 AM)
Quoting : "Half the files are under 10MB"
Chris Brew: (10:09 AM)
CMS does the same
Jens Jensen: (10:12 AM)
https://indico.cern.ch/event/265334/
Duncan Rand: (10:13 AM)
the log files on the panda page seem about 0.5 MB
Jens Jensen: (10:13 AM)
http://indico.cern.ch/event/289680/contribution/5/material/slides/
Duncan Rand: (10:24 AM)
is my sound OK by the way?
Ewan Mac Mahon: (10:24 AM)
I could hear you 'OK', FWIW, 'clearly' would be slightly overstating it.
Duncan Rand: (10:25 AM)
OK is OK I think given I'm on a 1.5 Mbps link..
Christopher John Walker: (10:25 AM)
You are quiet and a bit croaky. 
but audible
Duncan Rand: (10:25 AM)
Chris: any idea where this comes from: 
Issue with MTU setting. Suggestion for LHCONE is to use jumbo frames. We need to understand the impact on our measurements and our infrastructure. 
Jens Jensen: (10:33 AM)
http://ec.europa.eu/research/participants/portal/desktop/en/opportunities/h2020/topics/2141-einfra-7-2014.html
Ewan Mac Mahon: (10:35 AM)
Our whole infrastructure may not be perfect, but it is hands down streets ahead of anything else.
I don't think anyone else has anything with the same combination of federated auth with delegatable credentials, and they're both critical properties to making all the good sruff work.
Duncan Rand: (10:39 AM)
eudat?
Ewan Mac Mahon: (10:46 AM)
Sanger's embassy system is very much about the bandwidth though.
They have to trust the VMs, and the companies (somewhat) have to trust Sanger.
raul: (10:47 AM)
you might need to store/search/process with encrypted data
Ewan Mac Mahon: (10:49 AM)
Problem with encryption is that to access it you have to give the end site the decryption key - this is the standard DRM type problem.
Unless you can process the data while it's still encrypted.
raul: (10:51 AM)
yes. process while still encrypted
homomorphic encryption
I had the idea thatf Sanger did that
Ewan Mac Mahon: (10:51 AM)
Ah, well that's exciting stuff, but last I heard that was still the domain of pure maths research. I think we're a way of deploying it.
raul: (10:52 AM)
yes. we don't do it
Ewan Mac Mahon: (10:53 AM)
I think we're pretty good on having data and not allowing the wrong people to access it.
What we can't do is control what the 'right' people do with it once they've got access.
And there's no defense against sysadmins.
Christopher John Walker: (10:53 AM)
Hydra was what I was thinking about