Enabling Grids for E-sciencE Batch System Integration Jan Just Keijser Nikhef Amsterdam EGEE III SA3 Kickoff Meeting, May
2 Outline Targets/Milestones Work done so far Next steps Issues > Other Interesting System Integration Issues
EGEE III SA3 Kickoff Meeting, May Targets & Milestones First milestone Take part in writing MSA2.3, 'Strategy and roadmap of the EGEE multi-platform support’ Near-Future targets Full support for 2 extra batch systems in gLite (apart from PBS and LSF) Both lcg-CE and CREAM-CE supported Develop a Batch System Integration Checklist (BaSIC ;-)) Write a 'cookbook' How to integrate Your Favourite Batch Systems in 6 Easy Steps
EGEE III SA3 Kickoff Meeting, May Work done so far... PIC: –Condor integration with lcg-CE: see Christian’s talk –CREAM-CE integration in the works CESGA: –Sun Grid Engine (SGE) integration with lcg-CE –SGE already certified by SA3 minor note: info-dynamic-scheduler ? Nikhef: –Inventorized existing batch system integration for Condor, SGE, PBS –Set up testbeds for off-developer-site testing of BSI CentOS 5.1 host OS, Xen 3.2, LVM Testbed consists of Batch System + WNs, CE, RB/WMS, MON, UI
EGEE III SA3 Kickoff Meeting, May Inventorizing... EGEE-UF03 in Clermont-Ferrand was very useful:
EGEE III SA3 Kickoff Meeting, May Next steps... At Nikhef –we will set up a testbed with Batch System Of Choice –Next, we will use YOUR installation guidelines/Twiki pages on how to integrate this batch system with gLite What we need from you: –installation packages –installation guides/Twiki pages –Time & effort –your cooperation
EGEE III SA3 Kickoff Meeting, May Longer-term ideas Possible ARC integration –ARC-glite interface is at CE-level –Thus, ARC integration could be treated as 'Just Another Batch System' Other batch systems –?? Other platforms –32bit Linux flavours, RPM-based,.deb-based –64bit Linux flavours, RPM-based,.deb-based –Others??
EGEE III SA3 Kickoff Meeting, May Issues Which site will develop and support LSF integration? Interaction with JRA1 developers is limited thus far –Who will/should own the interface between a batch system and the affected gLite components? –How to address issues like the current blahp interface issue, where PBS+LSF are treated differently from Condor CREAM-CE is very much a “moving target” Mismatch between FTEs available for doing coordination and FTEs available for technical/development work
EGEE III SA3 Kickoff Meeting, May Concluding Remark To make Batch System Integration a success it is vital that we have stable interfaces between the Batch System and the glite Middleware and that these are “owned” by SA3 Development JRA1 Batch system co-ordination Nikhef, SA3 Operations SA1 SGE CESGA,... LSF Condor PIC Torque/Maui Nikhef Batch system integration