Tuesday, January 12, 2010

The LHC Computing Grid in the starting blocks

Talking with Ian Bird, from the CERN Bulletin.

As the Large Hadron Collider ramps up operations and breaks world records, it is an exciting time for everyone at CERN. To get the computing perspective, the Bulletin this week caught up with Ian Bird, leader of the
Worldwide LHC Computing Grid (WLCG). He is confident that everything is ready for the first data.

The Worldwide LHC Computing Grid (WLCG) collaboration has been in place since 2001 and for the past several years it has continually run the workloads for the experiments as part of their preparations for LHC data taking. So far, the numerous and massive simulations of the full chain of reconstruction and analysis software could only be carried out using Monte Carlo simulated data. Now, for the first time, the system is starting to work with real data and with many simultaneous users accessing them from all around the world.

“During the 2009 large-scale computing challenge (STEP’09), several major milestones were achieved: first the data transfer rates sustained were well above what was actually designed for – we achieved sustained aggregate data rates close to 4 GB/s – more than twice that required. This is equivalent to transferring a DVD of information every second. Secondly, the Tier 1 sites were able to show that they could accept this data stream, archive it on tape and simultaneously recall data for processing – all at the rates required during full scale LHC running, and in many cases well in excess of that rate. Finally, but perhaps more significantly, the experiments were able to demonstrate that the system could support large numbers of users running ‘real’ physics analyses on the data”, says Ian Bird.

Of course, the experiments have actually been collecting real data from cosmic rays for the past several months (not to mention some real collision data in the last weeks!) and have been putting the entire grid system through its paces so as to align and calibrate their detectors in preparation for full-scale data taking. ”Not only do we believe the system is ready, but it is actually in daily use already!”, confirms Bird.

Despite all possible simulations and tests that can be performed, no system can be fully understood until it is in use for real. “Undoubtedly there will be surprises when we start handling real data and we have to be ready to react and adapt to those situations. While in 2009 great steps were made in supporting many more users than just the experts, we still have a long way to go. This will no doubt be one of the areas where we have to be ready to adapt and improve things. There are several developments in hand that should help in this area and deploying those will happen this year”, assures Ian Bird.

During the past several years WLCG has made use of several grid infrastructure projects, including EGEE in Europe and OSG in the USA. Now, the landscape of European Grids is changing because the EGEE will come to an end in April 2010 and a new structure based on National Grid Infrastructures with European level coordination will be put in place. “This is potentially a major change in the underlying support structures for WLCG and, of course, the timing of this – just as the LHC physics programme takes off – is unfortunate”, admits Ian Bird. “However, that is the way it is and we have to focus on making this transition as smooth as possible”.

At the time when such a big project reaches this phase, the feelings of people who have built it must be of real anticipation. “Certainly not all of the grand ideas that had been discussed have come to fruition, but on the other hand today we really do already support data rates and workloads well in excess of those originally planned”, says Ian Bird. “The worldwide collaboration in computing that we have built in WLCG is really a first and I think we can all be proud of that – this surely will serve as a model for other international science communities that are on the horizon that will eventually produce data in amounts that will dwarf what LHC will produce, but we did it first! But now it really is time to put this to the test with real data ...”

No comments:

Post a Comment