60 likes | 141 Views
Joint Meeting of the US, AUS, and XS Working Groups. TG09 Tuesday June 23, 2009 1330-1530 hrs Potomac 5/6. Agenda. Who are we and what do we do? Frontline, advanced and extreme-scalability user support tasks defined (Sergiu, Amit, Nick) What are the best practices for…?
E N D
Joint Meeting of the US, AUS, and XS Working Groups TG09 Tuesday June 23, 2009 1330-1530 hrs Potomac 5/6
Agenda • Who are we and what do we do? • Frontline, advanced and extreme-scalability user support tasks defined (Sergiu, Amit, Nick) • What are the best practices for…? • Porting and debugging codes • Dealing with “unforeseen” runtime errors • Optimizing and scaling codes • Recording lessons learned and bootstrapping documentation and training • Understanding and assisting new users, communities, and applications
Frontline Support Tasks • Prompt and successful resolution of complex user problems • For issues specific to 1 RP, share best practices and salient lessons across all RPs • Form tiger teams to diagnose and handle cross-RP issues • User Engagement • Make every user contact a mutual learning experience • Feed user concerns, experiences and suggestions into the TeraGrid management and infrastructure system • Be the users’ advocates and agents in the TeraGrid organization • Coordinate user surveys and other formal instruments
Advanced Support Tasks • ASTA (Advanced Support for TeraGrid Applications) • Guided by the allocations process (TRAC, Startup, Supplemental) • Consider - AUS staff expertise, RP sites of resource, work plan, collaboration from PIs - when deciding on ASTAs • ASTA primarily involves work with one user’s code • ASP (Advanced Support Projects) • Install, maintain domain science, HPC software, codes • Work on technical projects, identified with input from users by AUS staff, that can impact many users/groups • Document exemplary use cases • ASEOT (Advanced Support for EOT) • HPC/CI training, outreach (DataNet, PlantCI, other NSF directorates), workshops/tutorials (TG09, PetaScale etc.)
Extreme Scalability Tasks • RP staff, tool developers, and users work together to address issues that manifest at extreme scale • Scalability and Architecture • algorithms, fault tolerance and resilience, numerical stability and convergence • hybrid/multicore issues • Tools • performance tools, debuggers, compilers • focus on applying tools to production applications at scale • Workflows, data transport, analysis, visualization, and storage • data movement, parallel I/O, data-intensive analysis