OWG Meeting 5-May-2016

From GlueXWiki
Jump to: navigation, search

Location and Time

Room: CC A110

Time: 2:00pm-3:00pm

Connection

You can connect using BlueJeans Video conferencing (ID: 120 390 084). (Click "Expand" to the right for details -->):

(if problems, call phone in conference room: 757-269-6460)

  1. To join via Polycom room system go to the IP Address: 199.48.152.152 (bjn.vc) and enter the meeting ID: 120390084.
  2. To join via a Web Browser, go to the page https://bluejeans.com/120390084.
  3. To join via phone, use one of the following numbers and the Conference ID: 120390084
    • US or Canada: +1 408 740 7256 or
    • US or Canada: +1 888 240 2560
  4. More information on connecting to bluejeans is available.


Previous Meeting

Agenda

  1. RAID disk upgrade
    • Increased capacity (72hr*800MB/s = 207TB > 144TB (100TB) )
    • 3rd RAID
  2. Announcements
    • New meeting time = 2pm
    • gluon01: RHEL7-> RHEL6
      • gluon46 will remain at RHEL7
  3. Collaboration Meeting Agenda
  4. Compiler upgrade (gcc 4.9.2)
    • sim-recon will require it starting June 1st
    • ROOT -> v6 (not immediate, but will likely follow
  5. DAQ
  6. L3 Status
    • Rates plot
    • Farm upgrade:
      • 8-10 new nodes (~2x as fast as current 12 farm nodes)
      • If L3 node runs at 1kHz, we'll need equivalent of 100 nodes
  7. ROL status (SYNC events)
    • Scaler Data format?
  8. AOT

Minutes

Attendees: David L., Sergey F., Paul L., Hovanes E., Curtis M., Sean D., Beni Z., Simon T., Eugene C.

RAID Disk Upgrade

  • David summarized some points of a conversation he had with Chip and Sandy the day before
    • With our current 800MB/s data rate, we no longer have a 72 hr buffer of disk space
    • Our existing servers may be coming to the end of their warranty period
  • Currently, we always keep some data on RAID disks and so only tend to free up 50TB when a disk is cleared out
  • This makes our effective current buffer ~100TB = 34 hrs of buffer
  • Existing drives in gluonraid1 and gluonraid2 are 3TB. Replacing them with 6TB drives would approach cost of new server with new drives
  • New server similar to our existing one but with 6TB drives would be an estimated $14k
  • Computer Center has started using 3rd party company service contracts for hardware to replace expiring warranties
    • Our existing raid servers will hit 3 years this Sept. and we should look at setting up such a service contract from ~$100/month
    • Paul is looking into identifying RAM for us to purchase to increase from 32GB to 64GB in gluonraid1 and gluonraid2
  • Proposal for 3rd RAID as small buffer was presented
    • Addresses issue of stopping and restarting DAQ so ER can be run on different node when switching disks
    • Can limit access to disk from other users while DAQ is running, reducing potential conflicts
    • We decided to explore such a scenario using the existing 2 RAIDs and the 3rd RAID described above.

Announcements

  • We will change the meeting time to 2pm, but keep it on Wednesdays
    • Hovanes noted that this can interfere with the Wed. afternoon Accelerator meeting
    • We decided to keep it on Wed. since we don't usually have Online Meetings during the run anyway due to daily RC meetings.
  • Paul will likely do the gluon01 downgrade next week. We told him there was no time pressure since people will be occupied with the Collaboration Meeting/Workshop

Collaboration Meeting Agenda

  • Agenda is now frozen. Naomi will speak for precisely 10 minutes. No more, no less.

Compiler Upgrade

  • The offline group will officially require a C++11 compliant compiler (e.g. gcc 4.9.2) starting June 1st.
  • The hdops account uses sim-recon for monitoring histograms and hdview2 so we will need to support this change
  • Hovanes argued that making the default compiler something other than the current 4.4.7 would require significant effort to not just recompile all of the controls software, but to test it all and ensure it still works correctly
    • Controls group has full list of high priority tasks prior to the Fall run with no significant manpower available for this testing
  • If the default compiler for hdops is not changed to gcc 4.9.2, then programs compiled with it will need to be outfitted with launch scripts to setup the proper environment before the program is run
  • No consensus was reached. Further debate was deferred to a later date

DAQ

  • Sergey had to leave so no DAQ report

L3 Status

  • New low-level EVIO parsing code is still under development, but seems to be working at roughly 3x the speed of old parser
  • Lower level reconstructed objects (e.g. DTrackCandidate) are produced with the same parameters, but higher level ones (e.g. DTrackTimeBased) have some discrepancies
  • Current best estimate is L3 will be able to process 1kHz per node (or equivalent thereof)
  • New nodes (to be purchased this year) should be about 2x as fast. Thus, current estimate is we will need 50 of those types of nodes for L3
  • Will be purchasing 8 new nodes as part of CC annual procurement

ROL Status (SYNC events)

  • Alex S. not at meeting so nothing was reported