CODA

1
CODA

• Graham Heyes
• Computer Center Director
• Data Acquisition Support group leader

2
About CODA

• Common DAQ system at Jlab
• Designed to be
• Extensible.
• Modular.
• High speed.
• Tailored for Nuclear Physics DAQ.
• Supported by JLab staff.
• Design driven by Jlab requirements.

3
GlueX requirements

• Rates
• Event rate at L1 200KHz
• Event size at L1 5 Kbytes
• 1GByte/S off the detector.
• 15 to 20 KHz after L3 100MBy/S to storage.

4
CODA 2003

• Current front-end hardware
• Motorola PPC processors.
• Hall B 20 Readout Controllers in parallel.
• Highest event rate 10KHz
• Highest data rate 20MByte/S to disk.
• Event Builder
• Switched Gigabit ethernet.
• UNIX (solaris or Linux)
• Multithreaded on SMP hardware.

5
Limitations

• Hardware interrupt or polling limits event rate.
  Depends on CPU speed and architecture.
• Network protocol efficiency limits event transfer
  rate.
• Throughput of event builder and speed of disk
  drives limits throughput to disk.

6
Solutions

• Trigger rate too high for per-event readout
• Read blocks of events
• Requires hardware tagging with event identifier
  (number).
• 1KHz block rate for 100 events per block is
  similar to current rates in halls A/B/C.
• Data rate too high for single data link or CPU.
• Parallel readout at front end.
• gt 20 readout controllers can handle 1 Gbyte/S
  from detector.
• Parallel event builder.
• Number of event builders depends on technology
• gt 10 EB for current 1GBit/S networks.
• Parallel streams in online L3 farm.

7
Readout Controller

• Read blocks from detector at 1KHz.
• Reformat data blocks into events.
• Arrange events into blocks for transmission.
• Transmit event blocks to Event Builders.
• Use ET system to transfer data.
• ET handles load balance and fault tolerance.

8
(No Transcript)
9
Control and Monitoring

• Current control system for CODA is slow and hard
  to maintain.
• New control system, written in Java for
  portability is being written.
• New system will incorporate a slow control
  system.
• New system will integrate with EPICS.
• Easier to use than EPICS.
• Designed for rapid design and prototyping.
• Use industry recognized communication standards.
• Will allow CODA to compete with lab view and
  other integrated DAQ software.

10
Summary

• Rates can be handled by existing technology.
• Task gets easier and requires less hardware as
  technology advances.
• Problem areas
• Storage is not advancing as rapidly as required.
• L3 farm needs to be there close to startup.