OWG Meeting 16-Jan-2013
- Still need to arrange for Ed to talk more on FPGA programming
- Review of minutes from 19-Dec-2012 meeting
- Hall D account strategy - All
- Online build system and code management - Elliott
- Level 3 farm requirements - Dave L
- Status of simulated raw data generation and analysis - Dave L, Dave A, Elliott
- Counting House furniture - Hovanes
- Solenoid test update - Elliott
- EPICS - Hovanes
- Controls - Elliott, Hovanes
- JInventory database - Elliott
- DAQ group projects- Bryan, Dave A, Graham, Carl, Vardan, Elliott
- EVIO, cMsg, CODA3, ECS/CSS, disentangler, readout list, Java/cMsg GUI
- Electronics, trigger and DAQ status reports - Chris, Alex, Dave, Fernando, Ben, Ed, William, Bryan
- Mantis Task Tracker
- Midas Elog]]
- Calibration/Alignment issues?
1:30 PM Wed 16-Jan-2013 CC F326
Note: we are skipping the meeting in two weeks due to interviews
1:30 PM Wed 13-Feb-2013 CC F326
Present: Elliott W, Bryan M, Simon T, Hovanes E, David L, Carl T, Eugene C, Dave A.
Hall D Accounts
See proposal in Wiki above. Everyone agreed to start with the simple strategy and use ssh to special accounts instead of passwords. Later Kelvin was happy with our strategy.
Online Build System
There was general agreement to start by testing Maurizio's SCONS system on sample online code. Elliott will give it a try with some of the farm node testing code he recently developed. We will stick with the Offline group's Make system for JANA-based software.
Level 3 Farm Requirements
See Dave's slides for full details. Dave currently achieves 22 Hz per core performing all the analysis in his slides. Optimization may decrease this, e.g. by using a staged approach and/or multi-dimensional techniques to allow a decision to be made before all the analysis is performed. Note that wire-based tracking takes 2/3 of the cpu time.
Simulated Raw Data
Dave L found misformed events, it is not clear if the problem is in his code or Dave Abbott's mc2coda package. Dave is almost done decoding mc2coda raw data, next up is mc2coda compressed format. Eventually the translation table will have to be implemented in JANA in a standard way to convert from crate/slot/channel to detector coordinates. Currently this project is not top priority (later...new data challenges and upcoming reviews may increase the priority of completing this project).
Counting House Furniture
Hovanes reported on meeting with Facilities Mgmt concerning outfitting the counting room and office area with furniture, consoles, displays, computers, etc. Further meetings are planned, but we have narrowed the choices for most items. The goal is to outfit the C.H. by the summer. We agreed that we don't need drawers or cabinets as part of the consoles as we want to maximize the amount of usable console, keyboard and monitor space. Running power to the control room is on hold until the console furniture is ordered.
Dave A. led a discussion on EMU requirements:
- EMU's do the most intensive computing and need the most powerful machines.
- 16-core machines should suffice, maybe 32-cores needed for second level event builder.
- Core speed not critical, but need lots of memory, 2 GB/core..
- First level event builders need both 10 GB and IB ports, remainder just need IB.
- Need to check board architecture and how I/O is managed.
- Expect max 10-12 ROCs per EMU ==> min 8 boxes.
- Need latest architecture CPU's, fast memory.
Carl reported that he speeded up the buffer i/o functionality, needed when dealing with events in ET.