DUNE UK Computing Meeting, 16 April 2019
========================================
https://indico.fnal.gov/event/20494/

Andrew McNab (AM, chair)
Peter Clarke (PC)
Chris Brew
Fergus Wilson
Jarek Nowak
Raja Nandakumar
James Perry

Introduction (AM)
============

Created Computing Consortium earlier this year. Leader Heidi Schellman. 
Host lab technical lead is Mike Kirby, and I am the international 
technical lead (in practice I am doing WLCG "process" integration and 
some workload management evaluations / prototyping.)

MC and protoDUNE data production work is being submitted to multiple 
grid sites. Quite bursty and manually driven.

IRIS Resources (PC)
==============

UK comp resources from IRIS (and named for DUNE) but GridPP also 
contributing. DUNE needs more than can get from IRIS. DUNE asked for 
"1000 cores" for next year: allocation is 500 from Tier-1 and 500 from 
rest of GridPP-managed IRIS resources. Need stronger computing model for 
this allocation process next year. ie a protoDUNE comp model covering 
now onwards. For storage, 1000TB at Tier-1 and 1000TB at rest of 
GridPP-managed IRIS. 500TB of tape.

Roundtable
==========

Edinburgh (Clarke, Perry)
---------

Effort to contribute to RUCIO development. 1+ FTE, GridPP and possibly 
IRIS funded. PPRP round post also requested for DUNE DM. James Perry 
shared with LSST. Initially tried to get list of things to do in RUCIO 
from DUNE. Not clear. Specific list from RUCIO core developers. 
Delivered core support for object storage (giving RUCIO ability to 
generate signed URLs to avoid client having full credential, with S3 and 
Swift). Next work is on deletions, and third party transfers, and after 
that ability for RUCIO to interface to external metadata catalogues.

Teng is working on RUCIO monitoring.

Need to make sure all this is joined up with RAL RUCIO work.

Lancaster (Nowak)
---------
Not part of Computing Consortium. Allowed DUNE VO to submit to grid 
resources at Lancaster. Investigating why jobs not running in recent 
productions etc.

Manchester (McNab)
----------
WLCG integration (GGUS etc). Setting up DIRAC evaluation (SAM file 
locations and metadata in DIRAC File Catalog for testing, demonstrated 
matching, wiki page explains how to use this with a FNAL credential); 
DUNE VMs for OpenStack, adapted from FNAL HEPcloud VMs.

RAL-PPD (Brew, Nandakumar, Wilson)
-------
Been trying to get data onto RAL Tier-1. Central production workload is 
bursty. Run single particle MC production for DUNE. Setting up more 
production. GPU farm at RAL-PPD, to provide capacitgy for machine 
learning development. Will allow people to log in and use.

AOB
---
Need to understand what is strategic and what is going to be long lived 
to work on. Running some productions on DIRAC would be a sensible one.

Will try to use Zoom for the next meeting.

Next meeting in 2-3 weeks time. (Will send a Doodle poll to the list.)