60 likes | 182 Vues
Action Items from TG Data Workshop. 11 January 2007 Kelly Gaither, JRay Scott, Phil Andrews, Mark Sheddon, Dane Skow. Action Items.
E N D
Action Items fromTG Data Workshop 11 January 2007 Kelly Gaither, JRay Scott, Phil Andrews, Mark Sheddon, Dane Skow
Action Items • USE CASE: Take user examples and document the process in a flow chart style. Identify possible tools used to accomplish the steps. Could start with GASOLINE and ENZO. • Sergiu to followup with consultant staff to understand the details of what people are doing today and where opportunities for • GASOLINE - PSC (Derek) • ENZO - SDSC (Robert and Lee ?) • USE CASE: Take the documented use cases and verify them through user services and the user community. • Sergiu will take these to the user community and work the verification discussions • The use cases have been updated in the data kit description in the Wiki • USECASE: Walk through in detail the use cases, look for simplifications and document the tools available to accomplish the tasks. • Part of HPC University development ? • Derek to look into this with Sergiu • ? Timeline complete in time to inform tutorial/focus group at TeraGrid 07
ALLOCATIONS: Need feedback and input to the data allocations policy document. • Dave Hart, Richard Moore, Craig Stewart to draft • PERFORMANCE: Arrange a presentation/documentation on capabilities and access protocols to TeraGrid Listener. • Matt/Tony to present to Data WG on timeline and capabilities • PERFORMANCE: Should have some place to post minimum performance and flag to the users that if they are not experiencing this performance, contact the help desk. • Need to created similar overview information in the user portal for data transfer capabilities • PERFORMANCE: Document and set the expectations for the user. For example, post past usage statistics which in effect imply what is likely to be possible in the future. Having just a single number is not ideal. • Start with analysis of GridFTP data. Best to provide information about actual user performance/experience • No consensus on what the user expectation is nor how to communicate this effectively • (No leader)
GRIDFTP RELIABILITY: Pursue naming convention for endpoints • JRay to followup on this with the data working • GRIDFTP RELIABILITY: TG Primer on TG services on the logging options and common usage. • Start with Preston and Michele’s study • GRIDFTP RELIABILITY: Have a data mover in theTG user portal. • Kelly following up on this with the User Portal staff • GRIDFTP RELIABILITY: FAQ for system administratorsand FAQ for users. • Knowledge Base section in WIKI today where people with draft KB articles should deposit them there • Matt will send info on how to contribute FAQ materials (separate for User and sys-admin facing information) • GRIDFTP RELIABILITY: Evolve Speedpage into composite of data from GridFTP Listener (real workload) and synthetic workload. • PSC team picking this up • GRIDFTP RELIABILITY: Should we investigate the KRB cross-site stuff between TG sites internally in parallel with the Globus work? What effort would that require? • Derek to work with Kelly for understanding possible scope and necessary team • GRIDFTP: Update the GridFTP servers • Dan and Kelly to work the discussion on Listener config and GridFTP upgrade cycle needs/plans (plan on Jan/Feb with current kit plans and then rapid turn in late Summer/early Fall • GRIDFTP: Review current log information for what performance data available • Preston, Michele to report to data working group on what they find by end of Feb. • GRIDFTP: JP will schedule discussion on Data-WG call about GridFTP configuration recommendation
GLOBAL FILE SYSTEM: General Strategy • Grow GPFS-WAN by exploring server re-export strategy • Watch Lustre-WAN options, expecting a tech rev which would make testbed viable for late 2007 • Dependent on the (Kerberos) security layer • Look for pNFS consolidation testbed possibility in 2008 • GLOBAL FILESYSTEM: Communicate with sites not deploying GPFS-WAN and find out why. • TACC, IU, Purdue and PSC present and spoke: license, platform support, reliability, usage cases • ORNL license cost • GLOBAL FILE SYSTEM: Organize meeting focused on exploring license options and costs. Address uncertainties. • Phil/SDSC to collect • Focus on re-export strategy to make this available • GLOBAL FILE SYSTEM: Develop guide on how to configure GPFS-WAN servers and re-export. Determine if viable means of alleviating architectural incompatibilities and possibly licensing issues. • Patricia, Michelle, Matt (UC ?) to estimate the effort needed to do this • Will work within the context of developing the CTSSv4 kit • GLOBAL FILE SYSTEM: Work on Caching demo with IBM for SC07 • SDSC (Phil), NCSA (Michele), +? • GLOBAL FILE SYSTEM: Need to deploy a team to look at data strategy moving forward as Track2’s and Track1’s come online. Will we get to a point where data stays in one place and all operations are performed at that location? • Noted as a technical strategy discussion needed. • Need a way to address this strategy discussion (Dedicated workshop for end 2007 ?)