[Halld-online] Hall-D Online meeting minutes

[David Lawrence]

Hi All,

 Minutes from yesterday’s Online meeting are not posted here and copied below.

https://halldweb1.jlab.org/wiki/index.php/OWG_Meeting_2-Dec-2015

Regards,
-David

Minutes[edit]

Attendees: David L. (chair), Simon T., Beni Z., Sean D., Curtis M., Dave A., Bryan M., Sergey F., Alex S., Mark I., Dmitry R., Hovanes E.
Announcements[edit]

A new Run Period "2015-12" has started
Some more space has been cleared out on gluonraid1 
A few TB are still in use for the pxi system used to record the data on the solenoid. David will look into having this automatically copied to tape. He and Beni will then look into what files (if any) should be deleted from gluonraid1
Sandy obtained a quote for a 12 bay (10 for payload) RAID disk to be used as a work disk in the counting house. This will initially provide us with 25-30TB of usable space if we set aside 2 disks as spares. This will also leave 2 bays open for future expansion.
Run Preparations[edit]

RCDB[edit]
An issue with the latest CODA inserting an "&" into the XML has been resolved and the RCDB system seems to be writing data into the DB properly
Dmitry has worked with Mark to identify a configuration issue with the web server they believe to be the cause for the RCDB web interface not working properly. They have passed this information on to Marty W. in IT who will have a look and implement the fix
Dmitry has been working on adding a search filter to the RCDB web interface to make it easier to jump to desired runs
Beginning of Run (BOR) record[edit]
No work yet. Sergey pointed out some complicating issues. We will continue to discuss this offline but it may not be implemented before the Spring run.
DAQ system development[edit]
Brief discussion on configuration options. Hovanes asked if we should have a GUI to allow users to control the settings kept in the configuration files. This was in response to a statement about the f125 modes having names in the documentation, but using numbers in the configuration package. Sergey noted that the plan is to give users only a limited number of overall choices for a complete set of run settings. Things deviating from that should be implemented only by experts.
ROL status (SYNC events)[edit]
Sascha has not been able to get the SYNC event code into the repository yet. It is still high on his list.
Some discussion on configuring modules of different type (f125 and f250) in single crate.
Bryan M. said there is a one line fix to allow this. He will send to Sacsha.
We discussed how we might incorporate generically handling multiple module types in the same crate so that can be put into our standard ROL. Dave and Bryan said it was possible and they might work on an example.
L3/event tagging[edit]
Some work has been done on benchmarking read/write speeds in order to identify bottlenecks in the data transfer in the counting house systems.
Some upper limit for RAM was observed at a 3-4 GB/s
This is actually dependent on the buffer size. 2-4kB buffers can achieve >19GB/s, but 10MB buffers can only transfer 3-4GB/s
When only reconstituting events and transferring data (i.e. no L3 algorithm is run) about 1/3 of the time is spent byte swapping.
All data in CODA is in big-endian format because that is native to JAVA
L3 process must byte swap event when reading in, but then byte swap again when writing out
Dave and Carl have discussed it and Carl was going to look into making the ER allow little endian data as input and output. (This may mostly be there already, but needs to be reviewed.)
Writing little endian to disk will save byte swapping every time the file is read in the offline. 
AOT[edit]
Beni is currently using roctof2 for a long run. If anyone needs it, please contact him so he can stop his job.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.jlab.org/pipermail/halld-online/attachments/20151203/8cdf71e7/attachment.html>