Choose timezone
Your profile timezone:
GridPP Technical Meeting
→
Europe/London
Virtual Only
Virtual Only
,
Description
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 the Technical meeting 9/12/2016
Discussion on hardware purchases.
For most Atlas sites it is clear whether or not they should buy storage. However there are some intermediate sites (Oxford, Liverpool etc) who should think about it carefully. It will be discussed at the Jamboree in January. However having the smaller sites diskless means that the range of technical solutions that we need to provide is much reduced.
LHCb are clear that what they need most is cpu for MC production.
CMS is not as advanced as Atlas but is looking into having diskless T2s. The line from Daniele was that if we have the storage in the UK and share the data across sites (with xrootd) then it is up to us where we house it. With this in mind Brunel is going to spend the money that it was planning to spend on cpu shortly but keep the money that it was planning to spend on storage to see how things work out. They will then either spend on disk or on more cpu if the UK is able to work in the way described.
There was a discussion about the usefulness of xroot caches and this is something that we should investigate.
The implications for supporting other communities should also be considered.
Also the extra burden that this will place on the big data hosting sites.
VAC
Introducing mixed sizes of VMs - hopefully released by end of year.
Andrew monitors Ganglia page for VAC (after question from Peter L.)
Problems with Liverpool and single core jobs.
Alice VM now supported.
Networking
ATLAS are testing IPv6 today to make sure that the whole chain works.
Security
There was a WLCG yesterday and there was work thread sharing and BRO intrusion detection system (networking traffic monitoring) . BRO can do things like shutdown different end points. Communication between entities is very important in this area so that sites can react before they are actually hit themselves . This can be done through very effectively MISP. This can potentially be coupled to BRO and this is being discussed in the working group. UK Groups should get in touch if they want to be involved in this. This should focused through the working group.
GridPP Dirac
Dune pushing 100K jobs through the system was interesting and has a higher than normal failure rate. However this appears to be at the site end as far as we can tell although it is still being investigated.
HEP Software foundation
Meeting San Diego early next year
AOB
This is the last meeting before Christmas so nadolig llawen a blwyddyn newydd dda i bawb (google can translate)
There are minutes attached to this event.
Show them.
