60 likes | 201 Views
Scientific Cluster Support Program Steering Committee August 25, 2003 SCS Project Team. Current Situation. Existing SCS Program 10 clusters in project, 4 year duration SCS project funded by overhead Clusters adhere to standard configuration
E N D
Scientific Cluster Support Program Steering Committee August 25, 2003 SCS Project Team
Current Situation • Existing SCS Program • 10 clusters in project, 4 year duration • SCS project funded by overhead • Clusters adhere to standard configuration • Only cost to scientific divisions is initial purchase • UNIX Group supports some clusters outside program • Yucca Mountain Project cluster 1 – 2002 • Yucca Mountain Project cluster 2 – April 2003 • Berkeley Center for Structural Biology – May 2003 • TB Structural Genomics – August 2003 • Interest rapidly growing • 32 node cluster – Phillip Geissler Material Sciences and UCB Chemistry Dept. • 10 node test cluster – Steve Louie/Jeff Neaton Molecular Foundry test cluster • Etc.
Success Leads to Challenge • Creating a viable business model • Issues to consider: • Drive down cost of cluster support • Make affordable to scientists • …..
Driving Down CostsLab-Wide • Current cost breakdown for overhead money • Staff – 75% of budget • Maintenance • Facilities • Staff cost only place to drive down costs • Cheaper staff – students, lower level CSEs • Fewer staff • Standardize for economies of scale • Issue: computer room space • Lower quality • “Fewer larger” rather than “many smaller” • ….
Making Affordable to Individual Scientist • What is affordable? • May drive provision of >1 service level • Options: • Recharge – scientist pays entire support cost • IMAP email model – overhead funds infrastructure, scientists pay incremental • Frame shop model – overhead funds infrastructure, scientists provide effort (science division postdoc supports) • ….
Issues Complicating Costs and Affordability • Computer room space • Security • ….