Large Hadron Collider Grid: one year on

By Jamie Shiers, LHC Grid.

Back to index page

Next article: News in brief

The LHC will generate an astonishing amount of data: about 15 petabytes every year. The problems caused by storing and processing this amount of data could only be solved with a grid solution. This led CERN to develop the Worldwide LHC Computing Grid (WLCG). The WLCG is separated into 11 Tier 1 sites and 120 Tier 2 sites. Tier 1 sites are large computer centres which are distributed around the globe and are responsible for storing the LHC data. Tier 2 sites process data made available by the Tier 1 sites.

Over the last year, the WLCG has used the LHC downtime to demonstrate complex use cases and seek out weaknesses in the production service. Whilst no substitute for real operation, these exercises have helped to establish a realistic view of what it means to run a service as complex as the LHC Grid. We have shown that we can manage the inevitable service changes (everything from new releases to security fixes) and handle service interruptions. In this respect, the Grid has lived up to one of its main challenges: the global service continues regardless of changes that occur to parts of the grid. In the view of the author, the WLCG is the most demanding and convincing demonstration of grid computing.

What of the future? It is to be expected that the pressures of real data taking, real processing and real analysis will throw up some surprises. Some things will work less well than foreseen, despite the extensive testing that has been performed. For example, patterns of data access - particularly from analysis jobs - may differ dramatically from predictions. However, innovation driven by necessity will give rise to pragmatic solutions to any problems that are found. Solutions that work well will be rapidly adopted, and those that do not work, will be discarded. LHC computing has only just begun.

http://www.tinyurl.com/yzp6xec