Difference between revisions of "EPSCI SRO Meeting Apr. 20, 2020"

From epsciwiki
Jump to navigation Jump to search
 
(One intermediate revision by the same user not shown)
Line 30: Line 30:
 
=== Minutes: ===
 
=== Minutes: ===
  
''TBD''
+
Attendees: David L., Carl T., Jie C., Nathan B., Vardan G., Chris L.
<!-- Attendees: David L., Carl T., Jie C., Nathan B., Vardan G., Chris L. -->
 
  
'''Announcements'''
+
'''EIC YR Meeting'''
* Teams Whiteboard feature does not seem to be present. David submitted a Service Now request and will follow up on it.
+
* Last Wed.  
** ''Update: This feature is not available in the "government version" of O365 at the moment.''
+
* Markus suggested possibility of doing all calibration/reconstruction online via LHCb
* Talks for upcoming EIC SRO workshop next month
+
** Some concerns were raised along with suggestions for consulting detector groups
** Everyone should think about what, if anything we can present
+
** Organizer referenced someone's calculation that indicated saving all EIC data would not be an issue
** ''Update: Looks like we may be too late since the agenda already is filled out. I also heard Chris say something about turning people away already. [https://www.jlab.org/indico/event/378/ https://www.jlab.org/indico/event/378/]''
+
** Vardan has started looking into detectors and found the situation is still very fluid
  
'''Hall-B Testbed'''
+
'''Stand alone TriDAS'''
* Ben fixed TCP stack on FPGA
+
* Chris has started looking into this. Code has been downloaded and dependencies are being gathered.
* Sergey B. asked folks to wait on using it until he has had a chance to test the fix
 
  
'''Offline TriDAS for testing'''
+
'''Hall-B testbeds'''
* Chris suggested setting up a purely software TriDAS instance that we could work with outside of the context of Hall-B
+
* Sergey B. has stood up another SRO test setup in Hall-B with 1 crate and 2fADCS
** Generally thought to be a good idea and Chris will look into doing it
+
* Able to supply ~500MB/s of noise
 +
* Uses clondaq3
 +
* Vardan logged in this morning to start playing with but was kicked off soon after
 +
* Vardan will try implementing a CLARA-based system that uses that system as input.
 +
* Setup on clonfarm1 in use by TriDAS team
  
'''ADIOS'''
+
'''ADIOS Testing'''
* Carl has continued his testing of the ADIOS I/O interface library
+
* Carl is communicating with the author of the higher speed engine for ADIOS that is currently running at only 1GB/s on a network that should support 10GB/s
* It supports multiple engines
+
* It seems our use case of very high bandwidth between two nodes may not be well matched to ADIOS's primary user base.
* The wide area network engine uses zmq
+
* If this does not turn out to be a simple configuration issue or rectifiable operator error, we may need to look for an alternative
** their system allows receivers to drop packets
 
** Carl's tests failed due to having too small of packets so the processes started and quit too quickly
 
* Started looking at SST engine
 
** sockets or RDMA (if it is available)
 
* Tested on Mac at 400MB/s
 
* Tested on INDRA 100Gbs system up to 1 GB/s (factor of 10 slower than hardware)
 
* Carl noted the ADIOS user interface is pretty simple and easy to use. The ADIOS authors have also been very response so far to questions/requests
 
  
'''WMS Meeting'''
+
'''WMS'''
* David suggested canceling the WMS meeting Fridays in an attempt to reduce the number of meetings
+
* We briefly discussed the status of the WMS development
* There were no objections
+
* Vardan has communicated with the TriDAS authors, but has not heard back on his latest questions which included an offer to collaborate
 +
* They may be bringing in additional authors to help with TriDAS rewrites and not have details of that effort worked out
 +
* Vardan will pursue testing CLARA as an SRO WMS (see above)
  
==== Action Items ====
+
'''DAQ specs from relevant experiments (EIC, SOLID, MOELLER, CLAS12, ...)
* Carl will work on increasing ADIOS bandwidth on INDRA
+
* Vardan is gathering a list to put into tabular form
* Vardan will work with Sergey B. to get access to the Hall-B testbed once it is available so he can start stressing the TriDAS system there
+
* Some estimate by Vardan already for EIC based on luminosity, cross-sections, detectors of ~25GB/s
* Chris will set up an offline TriDAS system so he can start testing it outside of the Hall-B environment
+
* Graham confirmed this was similar to what he recalled from last year in the semi-secret report (shhhhh...) if you exclude the vertex tracker
* Nathan will work on slides for the Marco SRO talk next week (Thurs. 4/23)
+
* Vardan will make electron version available somewhere

Latest revision as of 19:48, 20 April 2020

Connection Info:

This meeting scheduled via MS Teams. Join link should be in your calendar.


Agenda:

  1. Previous Meeting
  2. Announcements
  3. EIC Streaming Readout VI workshop
  4. Status of stand-alone TriDAS
  5. Use of Hall-B testbedS
  6. Status of ADIOS Testing
  7. Status of WMS design
  8. AOT

Useful References


Minutes:

Attendees: David L., Carl T., Jie C., Nathan B., Vardan G., Chris L.

EIC YR Meeting

  • Last Wed.
  • Markus suggested possibility of doing all calibration/reconstruction online via LHCb
    • Some concerns were raised along with suggestions for consulting detector groups
    • Organizer referenced someone's calculation that indicated saving all EIC data would not be an issue
    • Vardan has started looking into detectors and found the situation is still very fluid

Stand alone TriDAS

  • Chris has started looking into this. Code has been downloaded and dependencies are being gathered.

Hall-B testbeds

  • Sergey B. has stood up another SRO test setup in Hall-B with 1 crate and 2fADCS
  • Able to supply ~500MB/s of noise
  • Uses clondaq3
  • Vardan logged in this morning to start playing with but was kicked off soon after
  • Vardan will try implementing a CLARA-based system that uses that system as input.
  • Setup on clonfarm1 in use by TriDAS team

ADIOS Testing

  • Carl is communicating with the author of the higher speed engine for ADIOS that is currently running at only 1GB/s on a network that should support 10GB/s
  • It seems our use case of very high bandwidth between two nodes may not be well matched to ADIOS's primary user base.
  • If this does not turn out to be a simple configuration issue or rectifiable operator error, we may need to look for an alternative

WMS

  • We briefly discussed the status of the WMS development
  • Vardan has communicated with the TriDAS authors, but has not heard back on his latest questions which included an offer to collaborate
  • They may be bringing in additional authors to help with TriDAS rewrites and not have details of that effort worked out
  • Vardan will pursue testing CLARA as an SRO WMS (see above)

DAQ specs from relevant experiments (EIC, SOLID, MOELLER, CLAS12, ...)

  • Vardan is gathering a list to put into tabular form
  • Some estimate by Vardan already for EIC based on luminosity, cross-sections, detectors of ~25GB/s
  • Graham confirmed this was similar to what he recalled from last year in the semi-secret report (shhhhh...) if you exclude the vertex tracker
  • Vardan will make electron version available somewhere