Re: Notes, esp. CernAtSchool
> There is also the LUCID experiment which will launch
> on a satellite and start gathering data in low
> Earth orbit.
I'd like Liverpool to have a chance of processing some LUCID data.
I don't think we've supported any spacecraft here yet, unless it
was by accident.
> There are disadvantages of having multiple sites, that
> they are not always able to be equally responsive when they have to
> support many non-local VOs, and making sure that things work at many
> sites can be time consuming. However, this is the generic resources
> argument for "small" VOs, and not based on the wow-factor.
Liverpool supports 30 VOs already; one more will make no odds.
Also, we've been working to automate VO management as much as
possible (mainly due to the fact that we have so many).
> Tom needs to know more about meta-data and file management. Do we have
> a tutorial somewhere which is not out of date? Not so much what's in
> the files but more whether there are datasets and such. We did have
> guidance for non-LHC VOs, to be updated from time to time.
Ewan (and I) put up a "Grid User Crash Course" - I think it's still currentish.
Of course, it's very basic; just LFC and SRM usage.
https://www.gridpp.ac.uk/wiki/Grid_user_crash_course
There is other stuff too, such as https://www.gridpp.ac.uk/wiki/Data_Management,
but we could really do with beefing up our advice to new small VOs. At the
moment, all we do is explain the most basic toolsets, and leave them to it.
What I would like to see, documentation-wise, is this: a profile for a typical
new small-vo, with recommendations on suitable configuration procedures,
submission procedures and data procedures. Configuration procedures would cover VO
set-up and management, VO adoption by sites, UI set-up and management, user
set-up and management, on-going monitoring and administration. Submission procedures
would describe basic job control options, and at least one well-known
submission/job monitoring framework, preferably one that is well used. Data
procedures would describe basic data management options, and at least one
well-known data control and meteadata manemenmhgt rameowrk, prefreably one that
is well used.
That's a lot of work, so it is progressing piecemeal. For example, we recently
set-up instance UI etc.. Ultimately, when we've got all the blanks filled out,
it would be an integration job to pull it all together into the document
described.
Anyway, that's one idea.
Cheers
Steve
--
Steve Jones [log in to unmask]
System Administrator office: 220
High Energy Physics Division tel (int): 42334
Oliver Lodge Laboratory tel (ext): +44 (0)151 794 2334
University of Liverpool http://www.liv.ac.uk/physics/hep/
|