PHENIX DAQ Workshop January 3-4, 2002 Summary

Draft 1.1 January 10, 2002

John Haggerty, BNL

1. The performance goals are to push events from the DCM through the Event Builder to the ATP's at the highest speed consistent with the FEM's and multiplexed DCM's, which should be near 8 kHz. The goal of the data logging is to log data at 100 Mbyte/sec or more.
2. Many fundamental software components will be upgraded at the end of the run, as soon as possible. However, phoncs0 will remain intact with the final version of online software frozen as much as possible.
3. Right after the run with the detector still intact, several DAQ issues could be investigated if the power can be kept on:
   • the convert and transmission times
   • throughput of the event builder
4. Reliability of some of the FEM's should be investigated, particularly the MVD, TEC, and initial download of the EMC. Low Voltage Power Supply reliability should be studied and improved if possible.
5. Some FEM's require further development.
   • The five event buffer should be tested and fixed where necessary (PC)
   • EMC short format should be made to work and zero suppressed
   • The FEM CONVERT and ENDAT times should be optimized for all systems
   • The RICH FEM ARCNET complete initialization should speed up
6. Solaris, CORBA, and VxWorks will be upgraded to current versions. This will require many changes to running code to accomodate.
7. It may be feasible to offload from Solaris to Linux run control and its servers, however, the availability and stability of CORBA services must be investigated as a first step.
8. Diagnostics and messaging should be improved to allow more rapid diagnosis of problems.
9. It is desirable to speed up the startup and initialization of the entire DAQ as much as possible. If the upgrades to CORBA don't do this, the sources of the slowups should be investigated.
10. The EvBServer will be moved to the NT side in an attempt to speed up and parallelize the initialization of the Event Builder components, unless other upgrades obviate the need for this or make clear the origin of the slowness.
11. Additional DCM's and SEB's with JSEB's will be needed for 4 MUTR.N granulettes and 1 MUID.N for two new granules.
12. JSEB problems identified in this run will be studied: the "first event" problem, the need for padding at the end of many granules, the counters for multiple events per buffer, and the use of interrupts.
13. DSP code in the DCM's will be investigated to determine whether events can be transmitted faster than the presently measured 1.8 kHz; this may require re-coding critical code in assembler.
14. The Event Builder will be upgraded and expanded with additional SEB's, ATP's. It is expected that the ATM switch will be replaced by a Gigabit Ethernet switch if studies of a Gigabit switch confirm that the performance will be improved.
15. There will be alterations to the present network architecture to incorporate several switches managed by us.
16. There are several areas of R&D which may not be ready for deployment in Run 3 but which will be pursued:
    • an upgraded Partition Module which can create trigger primitives
    • ways for switching clocks which are less disruptive than what is done presently
    • compression of data in the ATP's or buffer boxes, among others