OSG Production Looking to the Future Dan Fraser

OSG Production: Looking to the Future Dan Fraser, OSG Production Coordinator OSG All Hands Meeting March 2011

OSG Usage Mar 2010 - 2011

What have we been doing? A dependable Operations center n Stable BDII configuration (and other services) Scaling to meet the increased demand n Fewer “middle of the night” crises Scaling the accounting system (Gratia) > 2 M peak transfers/day, > 800 K peak jobs/day Enabling HTPC across more systems n Currently ~8 systems (including CMS Tier-1)

OSG sites now enabling HTPC OU n The first site to run HTPC jobs on the OSG! Purdue Clemson Nebraska (several systems) San Diego SMU (Atlas Tier-3) FNAL (CMS Tier-1) Your site can be on this list!

Data scaling across the OSG

Reducing the Barrier to Entry for new VOs Attacking the VO “re-learning” problem Each new VO used to probe the “structure” of each system they want to use. ~12 mos to acquire the knowledge and begin to scale out Glide-in. WMS now the recommended model n OSG hosting a Glide-in. WMS factory for new VOs SBGrid is a model case n n Transitioned to Glide-in (~2 month wall clock) Scaled to ~7000 simultaneous jobs (no extra effort) Other VOs also starting to use this model n Glue. X, LSST, …

Where are we heading? Continued scalability of services n n Ticketing, Information Services, Discovery, Accounting, Dynamic Resource Provisioning … Top level BDIIs Removing the single-core job restriction n HTPC to be considered “normal” on a CE Will be enabled on all ITB systems Added to the baseline documentation for CE installations

Enabling Campus Usage Campus users need only a local identity Plan to offer an integrated package vs the cookbook model Use an existing Glide-in. WMS FE (submit host) for Campus users This single submission point can route jobs to heterogeneous clusters running different batch schedulers When a user obtains a credential, the model is the same for accessing the OSG Let’s talk if you are interested! fraser@anl. gov

Your thoughts …?