With the LHC scheduled to start up next year, Grid sites worldwide are ramping up their equipment and service levels in anticipation of a flood of data.
In the UK this summer, a team from GridPP, the UK particle physics Grid, completed a round of visits to 17 GridPP sites, aiming to assess how ready they are for the LHC data. Dave Britton, GridPP's project manager, co-ordinated the process, which ensured that each site would be ready for next year and found out how to deal with any difficulties. The personal visits allowed GridPP experts to talk to people at each centre in confidence. Despite initial scepticism, this proved a useful exercise, as it allowed sites to explain their priorities, constraints and concerns directly to the GridPP management.
Since March this year, GridPP review groups have toured each of the four regional centres – ScotGrid, NorthGrid, SouthGrid and London Tier-2 – searching for signs of problems and challenges to come. Tier-2s differ from their larger Tier-1 counterparts in that Tier-1s are primarily for reconstruction and long-term storage of LHC experiment data, while Tier-2s are typically used to produce simulated data and for data analysis by end users.
Although each site had a slightly different story, many of the same issues arose, and reviewers identified a mixture of good practice and areas where lessons could be learnt from other sites. The site visits unearthed several issues that the GridPP team is now addressing, including difficulties for smaller sites in meeting the two-hour response time required from Tier-2 sites, the complexity of installing new virtual organizations, and training in-site monitoring using Nagios and Ganglia software.
GridPP has already responded to many of the concerns raised. For example, the team has developed a policy on when sites should stop stalled jobs; has helped to ensure that security policies are harmonized so that the appropriate response to any security incident is clear; and is working on middleware support and release issues.