From: Bob Jones [Robert.Jones@cern.ch] Sent: Tuesday, October 01, 2002 5:25 PM To: WP Managers Subject: release planning Hi, Here is the latest version of the release planning that I will present tomorrow at the PTB. Cheers, Bob. - Assuming the show-stopper patches to Globus work, we will still need to move to the latest Globus release probably in October - need input from ITeam on this point According to the Iteam minutes this point was discussed at the last meeting but no dates were specified. Charles could you update us on this point? - How important is the VO-name insert into the proxy subject (since it requires patch to Globus code and change to mkgridmap and we expect VOMS to be available shortly afterwards) - check with Roberto C. - When could we expect to introduce Zamboni? Space management is going to be important for the data challenge continuations. To be discussed at a future meeting. - Nobody explicitly mentions slashgrid but everyone assumes it will be included - we need to clarify which WP is responsible for its introduction. To be discussed at a future meeting. - I'm waiting for dates for WP1 items. Expended when Francesco returns. Technical issues to be discussed in the future: - Space management: Zamboni does not solve all our problems (i.e. for RB, gatekeeper etc.) - Data management: Need to clarify our recommended approach to the applications - Use of slashgrid - Future platform support Cheers, Bob. 1. Fix "show-stoppers" for application groups - mware WPs (continuous) 2. Build EDG1.2.x with autobuild tools - Iteam Items 3 & 4 are the responsibility of the testing group. It very unlikely they will be fixed by the end of September. We need a date from the testing group on when the basic framework can be in operation. Based on this date they may move in the list 3 .Integrate testing framework and limited automatic tests with autobuild tools - testing group 4. Automatic installation & configuration procedure for pre-defined site - testing group 5. Start autobuild server for RH 7.2 and attempt build of release 1.2 - Yannick Patois (ready) 6. Updated LCFG - WP4 (Oct 1-4) New LCFG (LCFGng): most LCFG objects have been ported by WP4 but there may be a few exceptions Fabric monitoring agent and server (needs LCFGng) 7. Introduce RLS - WP2 (ready): Oct 7-9am 8. Introduce R-GMA - WP3 (ready): Oct 9pm-11 No security. No impact on WP1 & info. providers There is a 2 week gap in our scheduling between Oct 14-26. Most likely candidate here is new Globus release and testing the show-stopper problems are really fixed. To be determined by ITeam. 9. Update LCAS - WP4 : Oct 28-30 LCAS with dynamic plug-in modules - depends on edg-gatekeeper (better to wait for new Globus release?) Resource Management prototype - implies changes to batch system configuration and depends on GRAM (new Globus release?) 10. SE 1.0 - WP5 (end Oct): Nov 4-6 Needs RLS (WP2) SE 'service manager' will provide basic subset of SRM functionality Get and Put, possibly Copy For MSS, Get results in file being staged to disk Hierarchical directory structure possible even if MSS is flat SE filenames can be longer than underlying MSS Remote access to staged files using GridFTP (for CASTOR only) ACL control of access through SE using GACL 11. NetworkCost Function - WP7: Nov 7-8 Will publish to both MDS 7 R-GMA 12. GetSECosts - WP5: 11-12 Used by WP2 13. Introduce Reptor - WP2 (being tested): Nov 13-15 Needs SE 1.0 & GetSECosts (WP5) and NetworkCost function (WP7) 14. GridFTP server remote access to CASTOR - Castor group (end Oct) 15. Res. Broker - WP1 (dates?) Re-factor to reduce crashes and memory leaks from underlying SW layers. Provide hooks for dependent/ partitionable/ checkpointable jobs. Minimize persistent job info repositories 16. LCFGng for RH 7.2 - WP4 already tested with CERN RH 7.2 - but we what distribution will we use for the testbed? 17. GLUE modified info providers/consumers - WP1, 3 (GIN/GOUT changes) 4 (ready), 5 waiting for SE stuff to be finalised in GLUE schema Schedule after SC demos in November - i.e. once it has been shown to work Includes additions to R-GMA: - security - replicated registry and schema - canonical producer These are needed for the SE and the L&B. Items above this line constitute testbed 2 release 18. SE 2.0 - WP5 (end Nov) More MSS support - Lyon, RAL, SARA, (WP9, WP10) Tarballs - allows writing large number of small files to MSS in efficient manner Disk Cache management Direct GridFTP access to Castor (omit SE disk cache) 19. SE 3.0 - WP5 (end Dec) GridFTP done through SE Supporting arbitrary MSS ACLs applied to transfers too Date less certain - still investigating several server solutions RFIO through SE Posix I/O through SE should remove need for NFS with subsequent increased robustness, efficiency and allow true space reservation For the work of ATLAS, it is assumed the "show-stopper" fixes being tested now along with access to CASTOR will be sufficient. They need something stable by end of October which they can use from November onwards. MapCenter (WP7) : MapCenter does not need a independent slot for deployment over R-GMA. WP7 are working on integration with R-GMA, but MapCenter has no interface for other middleware (in particular, networkCost function does not use it). As a first step, continue to use transparently the new information system (GIN->RGMA>GOUT->MAPCENTER), and then will try to switch to query directly RGMA, but in any case, the only impact would be that the MapCenter Web Server could be down for a few days. Below are some further points from WP1 that need dates - I have added what I see as high priority elements: 1) MPI job submission (high priority) 2) User APIs (high priority) 3) Accounting infrastructure 4) Interactive job support (high priority) 5) GUI 6) Job logical checkpointing (high priority) 7) DAG submission (high priority) 8) Partitionable job support (high priority) 9) Advanced queries to L&B (Job Catalog) 10) LB - R-GMA integration (depends on when R-GMA will be deployed) 11) Full integration of cost estimation/accounting into scheduling policies 12) Resource reservation and co-allocation 13) Integration with WP2 Query Optimisation API 14) Zamboni (high priority)