GridPP Technical Meeting

Friday 27 Jan 2017, 11:00 → 14:00 Europe/London

Virtual Only

Andrew McNab (University of Manchester), David Colling (Imperial College Sci., Tech. & Med. (GB))

Fortnightly meeting for technical topics looking further ahead than the weekly ops meetings on Tuesdays. There are also dedicated storage group meeting on Wednesdays. Each topic can go beyond the nominal 5 minute slot whenever necessary.

Notes from Technical meeting.

 

Atlas 

Site smaller than RALPP, Liverpool, Oxford ECDF should not be buying  disk and site that are bigger should be buying storage. These four should decide on their own and will come up with different solutions. Durham are running running with an ARC cache of tens of TB and are no longer a small CPU site. We are already looking extra sites of volunteers. To use ARC caching something like a global file system is needed. There was some discussion as to whether or not a simple NFS would be performant enough and alternatives were discussed.

 

Small sites are in the situation where they have not enough effort to migrate to a new state that requires. There was some discussion about us having a team that go into smaller sites for a few days to help them change over. It was pointed out that each site will have its configuration system for an example. However over a few days these could be sorted out. It was likened to Grid-in-a-box

The role of documentation was also viewed as important and such a task force would take this into account.

CMS

Running a CMS site has always been a bit more heavy weight than Atlas. However running a diskless CMS site is quite straightforward and this has been done with Oxford. These have been pretty successful (see slides). We will extend this to have all data hosted UK site. Chris also showed that files were being used multiple times so xrootd caching would be helpful.

LHCb 

Would encourage sites to buy CPU rather than joining the group of T2 Ds. Looking at using xrootd (did look at ahttp but that has stopped at the moment). Starting to look at similar ways of working as CMS  described above. This will an evolutionary process.

 

Further discussion was put off until 10am next week.

11:00 → 11:30 What T2s should look like - what hardware and what technical services

• Atlas view
• CMS view
  -LHCb view

Speakers: Andrew McNab (University of Manchester), Chris Brew (CCLRC - RAL), Peter Love (Lancaster University (GB))

11:30 → 11:50 VAC status

Speakers: Andrew McNab (University of Manchester), Peter Love (Lancaster University (GB)), Stephen Jones (The University of Liverpool)

11:50 → 12:10 Using Containers

Speaker: Andrew David Lahiff (STFC - Rutherford Appleton Lab. (GB))

12:10 → 12:30 Catch up

• Storage
• Security
• networking

12:30 → 12:35 AoB