Choose timezone
Your profile timezone:
HEPiX Virtualisation Working Group Meeting
→
Europe/Zurich
Phone conference
Phone conference
Description
<sl>
Dial +41227676000 and enter access code 0186715 followed by #
Key ##1 to mute/unmute your phone
</sl>
HEPiX Virtulisation WG meeting minutes 2010-02-15
Present (in no particular order):
Tony Cass, CERN
James Adams, RAL
Ewan MacMahon, Oxford
Keith Chadwick, FNAL
Martin Bly, RAL
Sonder Klous, NIKEF
Ulrich Schwickerath, CERN
Ricardo Silva, CERN
Thomas Finnern, DESY
Andrea Chierici, INFN Bologna
Owen Synge, DESY
Ian Collier, RAL
Ian Gable University of Victoria
Stefan Roiser, CERN
Herman, Berlin (last name needed)
1. Tony gave a summary of the group's status to the GDB.
2. Image distribution.
There was quite a lengthy discussion in this area. I will attempt to summarize the keys points rather then capture all of it.
3. Contextualization
4. Hypervisors
5. The next meeting:
Present (in no particular order):
Tony Cass, CERN
James Adams, RAL
Ewan MacMahon, Oxford
Keith Chadwick, FNAL
Martin Bly, RAL
Sonder Klous, NIKEF
Ulrich Schwickerath, CERN
Ricardo Silva, CERN
Thomas Finnern, DESY
Andrea Chierici, INFN Bologna
Owen Synge, DESY
Ian Collier, RAL
Ian Gable University of Victoria
Stefan Roiser, CERN
Herman, Berlin (last name needed)
1. Tony gave a summary of the group's status to the GDB.
- The slides are linked from the agenda for this meeting. There were no objections to the slides.
2. Image distribution.
There was quite a lengthy discussion in this area. I will attempt to summarize the keys points rather then capture all of it.
- Owen led off the discussion, he pointed suggested that X.509 certificates should be used to sign certificates, there was no opposition to this point. There would appear to be consensus on this point.
- Owen also proposed that with signed images there is no need for a secure transfer protocol. There were no objections.
- Once the images are signed we need to have an authorization list of some kind.
- We then need some kind of catalogue to store the images and meta data about the images (version, expiry, hypervisor etc).
- The LFC was suggested as a possible catalog to use.
- Ewan suggested that images be pushed by VOs in the same way that VO software is now pushed. This triggered a discussion on if image should be pushed or pulled.
- Martin felt that it was not important to prescribe if the images were pushed or pulled as long as the image catalog held their locations.
- Tony will speak with Jean-Philippe Baud (LFC designer) on the suitability of the LFC fro our purposes.
- - Thomas Finnern commented that it would be important the the catalog be able to indicate replacement of images (image X replaces image Y) in order to avoid ever growing image repository.
- Owen commented that images are typically 2 to 4 GB without experiment software.
- Ewan pointed out that early testing will be important, he put forward his site, Tony also offered CERN.
- Clearly there is there is going to be more discussion on distribution required.
3. Contextualization
- Tony put forward the list of things to be contextualized as the following: site monitoring, site WMS, logging, scratch space, site file system.
- Ewan stated that he didn't see the point of contextualizing images to connect to local WMS, tony pointed out that this is what CERN is doing today, and that he agrees that in the future that images might connect to central schedulers, however this is up to the sites today.
- Owen put forward that the Kernel also be up for contextualization, to allow for increased performance in certain circumstances. In his experience there is little difficulty with replacing kernels within major versions of Scientific Linux.
- Ewan fears that contextualizing the kernel could introduce problems, with modules in particular.
- Further discussion on kernel contextualization of the kernel is required.
- Martin commented if we also need to contextualize data access? There isn't a clear answer here and it should be discussed at the next meeting.
4. Hypervisors
- Andrea commented on the doodle poll:
- Not sure if the poll is complete
- 14 respondents, consensus on KVM, Xen Fully Virtualized, and Xen para virtualized also popular.
- much less interest in other hypervisors.
5. The next meeting:
- Tony would the face to face meeting to be held on the Friday afternoon of HEPiX in Lisbon.
- The next phone meeting will take place March 1, 15:00 UTC.
- Topics identified for next meeting agenda:
- Andrea to discuss how KVM and Xen can be supported.
There are minutes attached to this event.
Show them.