From: Francesco Prelz [Francesco.Prelz@mi.infn.it]
Sent: Wednesday, January 16, 2002 2:41 PM
To: Olof Barring
Cc: Francesco Prelz; Francesco Giacomini; Massimo Sgaravatto; Maria
Barroso Lopez; bob.jones@cern.ch
Subject: Conclusions out of WP1/WP4 bi-lateral meeting.

Follow Up Flag: Follow up
Due By: Friday, January 18, 2002 5:00 PM
Flag Status: Flagged

This is just to formally agree on the conclusions of the meeting on
Monday (as they were minuted by Olof).

Let me recap the four WP1 issues:
 i  -  Will the gatekeeper provide network "tunneling" services? 
       Can we brainstorm for a little while as to how?
 ii -  What can we assume that the Y2 LRMS/gatekeeper/jobmanager will
       be able to do? (Perhaps this boils down to an explanation
       of the RMS design, which I'm not sure I could grasp from the
       Design document).
iii -  Where and how will the WP1 Accounting infrastructure and the WP4
       sensors meet ?
 iv -  What sort of CEs (in terms of advance schedule or preemption
       capabilities) will be available ? What information will 
       be published (if any) about the CE "schedule" ?

Here's my current understanding of the agreed status/answers:

  i -  FabNAT is not an IP tunnelling service, but just a regular NAT
       with the extra functionality of setting up direct inside-to-outside
       IP mapping through vendor-specific router access. This could become 
       useful when/if we need to build site spanning MPI support - it
       was agreed this is not foreseen to happen during Release 2. 
 ii -  No fancy feature of the "supported" local resource management systems 
       will be generalised within year 2, and made available through
       "GRAM". Hence the WP1 RB will have to do the necessary processing
       to transform multiple/dependent jobs into single submissions. 
iii -  We weren't able to really understand the on-going WP4 activities in
       the field of developing accounting sensors. WP4 will understand what
       they are going to do and how, will come back to WP1. We will then
       decide whether a meeting (in Paris) is needed with the WP1 Accounting 
       folks.
 iv -  The only possible "special" kind of queue that may be available
       in the Release 2 testbed is a quick-turnaround (pre-emptive ?) queue.
       This can be advertised as having a "zero" queue traversal time,
       and thus fits in the current CE objectclass schema. It looks like
       we don't expect changes in the CE schema over RElease 2.

In case WP4 agrees, we may be done with this, otherwise we'll go through 
another iteration.
Thank you for the helpful, co-operative attitude.

Francesco P.