LDR Meeting April 23, 2004

Logistics

• April 23, 2004
• 9:00 am to 4:00 pm
• MIT, NW22-220, Boston MA
• hotel information
• campus maps
• Keith has requested that Marie have the doors unlocked from 8-10am so there shouldn't be any problems. As a backup, please go next door to Marie Woods office NW17-161 (NW17 & NW22 are adjacent) and she can call us to bring over stragglers.

People

Confirmed attendees so far:

• Keith Bayer
• Kevin Flasch
• Mike Foster
• Ben Johnson
• Scott Koranda
• Brian Moe
• Hari Pulapaka
• Murail Ramsunder

Agenda

• 09:00 - 09:30 "LDR: Past, Present, Future" (Scott Koranda)
• 09:30 - 09:45 "Better Metadata for LDR" (Brian Moe)
• 09:45 - 10:00 "Plans for LDR Support" (Kevin Flasch)
• 10:00 - 10:30 "Leveraging the DiskCacheAPI from LDAS" (Hari Pulapaka)
• 10:30 - 10:45 break
• 10:45 - 11:00 "Making Python Go Faster" (Hari Pulapaka)
• 11:00 - 11:15 "Matlab, LDR, and Framequery" (Mike Foster)
• 11:15 - 12:00 LDR feature requests (group discussion)
• 12:00 - 01:00 lunch
• 01:00 - 02:00 requirements/plans for S4/S5 (group discussion)
  • diskcacheAPI and data discovery/publishing interaction with LDR
  • publishing interface/API?
  • diskcacheAPI as a LDR storage module?
  • single LDR interface/API?
• 02:00 - 03:00 prioritizing LDR development (group discussion)
• 03:00 - 03:15 break
• 03:15 - 03:50 draft plan document for LSC Comp Comm/DASWG
• 03:50 - 04:00 plan summer meeting in Europe (UK?)

Known LDR Feature Requests

• smooth upgrades with migration of metadata/RLS state
• choice to use exisiting MySQL server
• green/red light "at a glance" web page
  • gauge telling percentage of collections transferred
• global scheduling across multiple sites/collections (think LLO going down during S3)
• archived transfer peformance information
• scheduling based on archived transfer performance information
• on-the-fly configuration
• rotating log files
• validity checking at server and client end for transfers
• "proxy LDR"; if a LSCdataFind request fails that ups the priority for the failed file?
• Reliability of transfer features (I think 'validity checking' covers it).
• Optimizing transfer speed within reliability and security needs.
• Are we secure enough?
• Optimizing the LDR database for data analysis (think pounding!). The mySQL server can take it (typically loads of 8-10), but our experience is that it kills the RLS server and LDR and that is why we had to set up a separate db for analysis on another server. Is this problem solvable?
• (more coming...send to Scott Koranda)