Bill Appelbe bill@vpac , VPAC (vpac)

1. eScience Infrastructure and the Changing Culture of Research-Successes and Lessons Learned in Australia Bill Appelbe bill@vpac.org , VPAC (vpac.org) UK eScience Presentation

2. Outline • A Short History of eResearch in Australia • The Current Australian eResearch landscape • Future directions and lessons learned UK eScience Presentation

3. eResearch in Australia • HPC/Cyberinfrastructure in Australia is funded and organized very differently from the UK or USA • In Australia • There are only ~40 Universities (and 20m. people) • One national government research agency (CSIRO) • National research telecom – Aarnet.edu • Since 2005, focus on National Collaboration rather than competition for Research Infrastructure • So no “competitive bids” for Peak Computing facilities UK eScience Presentation

4. Evolution of Australian eResearch E I F NCRIS-1 APAC pre-APAC 1995 2000 2005 2010 • National research infrastructure collaboration & funding, including Advanced Computing (PfC) • Community-based • Maturing grid computing • SuperScience Initiative • State investment –BRC, MASSIVE, VLSCI • National HPC collaboration & funding: National HPC Tier-1 Facility (ANU) & state-based Tier-2 HPC facilities (PACs) • Merit-allocation for Tier-1 HPC • Incipient grid portals & operations • Universities purchased and supported their own HPC; little collaboration • Growing concern that Australia was falling behind UK eScience Presentation

5. eResearch in Australia – APAC Era • APAC, the Australian Partnership for Advanced Computing, was formed in 2000 • Ambitious idea • A national Tier-1 system at the Australian National University in Canberra • Commonwealth funding to States to set up and/or support State-based Tier-2 HPC Centres • $6M to Victoria to set up VPAC • Matched by State funds and University subscriptions UK eScience Presentation

6. VPAC – “Innovation powered by Advanced Computing” • VPAC is an Advanced Computing/Cyberinfrastructure “research services” organization • Like SDSC, NCSA, EPCC… • But organized and funded differently, and • Does not do independent research! • Independent company; owned by the State’s Universities • Collaborative R&D with Universities & government agencies • Such as Geosciences Australia (~USGS) and Geosciences Victoria • Also provides services to companies such as GM, Boeing, … • Now operates most HPC facilities in the State • But it is not “socialist” HPC! No mandate for anyone to use VPAC or State subsidies UK eScience Presentation

7. APAC Successes and Lessons APAC Successes • National collaboration, national merit allocation of HPC • Acted as a catalyst for formation of regional HPC Centres • Use of gateway machines (VMs) to support grid computing UK eScience Presentation

8. APAC Successes and Lessons Lessons • Putting $ into Universities to develop HPC courseware was not a success • Grid linking of sites did not generate a lot of use • As there was not a lot of “resources” (HPC cycles) put in; • And each State HPC system was individually managed; • And users do not like moving their applications and data • Grid portals were not very successful • Each individually funded, developed by grad. Students, not sustainable or maintainable UK eScience Presentation

9. APAC’s successors – 2007+ • NCRIS – research infrastructure funding • Including software development and funding • Programs in national priority areas such as geosciences, biosecurity, …. • Within NCRIS, an eResearch Infrastructure program • Plaforms for Collaboration www.pfc.org.au • Three subprograms: • ANDS – Data standards www.ands.org.au • NCI – Compute www.nci.org.au • ARCS – Australian Research Collaboration Services www.arcs.org.au UK eScience Presentation

10. ARCS Organization and Programs • ARCS was set up as an Unincorporated Joint Venture (UJV) of the regional HPC/eResearch Centres (the “MARCs”) • Staff distributed across the country/MARCs UK eScience Presentation

11. ARCS/NCRIS Successes Successes • Within NCRIS, ongoing funding for national software platform development • E.g., StGermain open-source computational science platform, used for geodynamics in AU, USA • In ARCS, development of • Cloud computing and configurable web portals: grisu • Data services, the “national drop box” • Collaboration tools (national EVO support) • National authentication/id • ARCS/NEAT Funding for community software projects UK eScience Presentation

12. ARCS/NCRIS Lessons Lessons • Organizations that are UJVs have governance and management problems • Not just ARCS, but also the VeRSI UJV, the Victorian eResearch Strategic Initiative • Engagement is critical to the success of an eResearch Service provider • With users • With University administration • Managing highly distributed software development teams is problematic UK eScience Presentation

13. Outline • A Short History of eScience in Australia • The Current Australian eScience landscape • Future directions and lessons learned UK eScience Presentation

14. Cyberinfrastructure in Australia(cont.) • VPAC has broad funding from Universities, state and federal grants, industry, etc. • ~$7M p.a. • 70+ employees at 4 sites in the State • Systems support, software engineers, engineers, scientists • Operates HPC and data centers for researchers and industry • 600+ users across 8 Universities; 5 HPC clusters; GPGPU and viz. systems • Professional software development teams for both academic and commercial projects, including computational scientists • VPAC staff “embedded” in Universities; joint grant proposals • Strong international links and collaboration • Joint software development for Geodynamics with USA since 2003 • The NSF Center for Computational Infrastructure for Geodynamics www.geodynamics.org UK eScience Presentation

15. Cyberinfrastructure in Australia(cont.) • As a state, Victoria has a mature but rapidly increasing Cyberinfrastructure investments • VLSCI - $100M joint investment by the State and Melbourne University in a Computational Life Sciences Center • MASSIVE - $10M investment in a Computational Imaging and Visualization Center by a consortium including the Australian Synchrotron, CSIRO, Monash, and VPAC • A $100M Biotechnology Research Centre • This complements the new EIF national investments • EIF Data Fabric (see .pdf attached) • Most Universities have “eResearch Directors” • Under PVCRs, facilitate eResearch access for researchers UK eScience Presentation

16. Outline • A Short History of eScience in Australia • The Current Australian eScience landscape • Future directions and lessons learned UK eScience Presentation

17. eResearch Services Centres • The theme of the workshop is “Advancing Computational Science in academia and HPC Centres” • I’m going to answer a more general question“What should you do to create a successful eResearch Services Centre” • eResearch services include computational science • But not everyone needs computational science, it is a “tool” just like “database design” • By implication, solving the more general question solves the simpler one UK eScience Presentation

18. eResearch Services Centres (cont.) Q: What does “success” mean in the context of an eResearch Services Centre? A: Common measures include • Sustainability and “critical mass” • Quality of services, or “value” to host/client research institution(s) • Innovation and strategic impact • Engagement – with users, host/client research institutions; HPC vendors; national and international collaborators UK eScience Presentation

19. Lesson #1 – Engage the user community • What are the research problems being addressed? • What resources are needed to tackle them? • Can the Centre help? How? Key failures and pitfalls in engagement: • Focus on IT first - “build it and they will come” • Inflexible project management • Focusing either on just the researchers or academic administration; or only high-end users • Communication: IT specialists and researchers do not mix UK eScience Presentation

20. Lesson #2 – Build and retain expertise • An eResearch Centre’s core asset is its expertise, not its hardware • You need both “breadth and depth”, and a culture of collaboration, not prima donnas • Train your own staff, and get them to train others Key failures and pitfalls in building expertise: • The NIH syndrome; or “solutions looking for problems” • Insufficient breadth/depth • Getting carried away with technology or stuck in a rut • Insufficient “outreach” or “eResearch Analyst” expertise UK eScience Presentation

21. Lesson #3 – Get the organizational structure right! • You need leaders who are both scientists and good managers • You need to be agile and able to redeploy expertise (“matrix management”) Key failures and pitfalls in organizational structure: • No governance Board, review, or oversight • No risk management/mitigation • Organizations too tied to, or hampered by, University politics, rules, and regulations UK eScience Presentation

22. Lesson #4 – Collaboration • Build meaningful, mutually beneficial ties and alliances with • Other Centres: regional, national, international • Industry: HPC vendors • IT development communities • Standards organizations • Government Key failures and pitfalls in collaboration: • Investing resources in “one sided” collaboration • Focus on “marketing” not “collaboration” UK eScience Presentation

23. Lesson #5 – Grow to a sustainable size • At least 40 to 50 technical staff is ideal, or • If smaller specialize and outsource skills through collaboration • Diversified funding Key failures and pitfalls in organizational size: • Over a dozen staff requires experienced managers • Researchers are notoriously bad managers • Not planning for growth in HR, finance management, or project tracking • Insufficient funds for organizational size or lack of discretionary funds UK eScience Presentation

24. The “old model” of HPC Centres • Focus on HPC and “big iron” • Users are expert UNIX users and programmers • Projects are small specialist research teams • Coding from scratch – “hero codes” • Data is secondary • Staff are specialist systems administrators and software developers • Limited in-house training UK eScience Presentation

25. The “new model” of HPC Centres • Focus on grid computing, diverse Advanced Computing infrastructure • Emerging users are not “traditional scientists” • Projects are diverse, collaborative, involve industry or government collaborators • Community codes; commercial software • Data may be primary (e.g., Biogrid, Biobank) • Staff include outreach experts (“missionaries”) with scientific background • Training and building skills is key UK eScience Presentation

26. Thank You! Questions? UK eScience Presentation