Hall D Trigger and Data Rates

1
Hall D Trigger and Data Rates

• Elliott Wolin
• Hall D Electronics Review
• Jefferson Lab
• 23-Jul-2003

2
Outline

1. Rates from Design Report
2. Comparison with LHC,CLAS
3. Additional Considerations
4. DAQ Challenges

3
1. Rates from Design Report

• High trigger rate 200 KHz
• Deadtimeless, pipelined front ends
• Small event size 5 KB
• Small Level 1 rejection rate factor of 2
• Modest rate off detector 1 GB/sec
• Modest Level 3 rejection factor of 10
• Modest cpu needed in Level 3 0.1 SPECint
• High rate to tape 100 MB/sec

4
1. Rates, cont
5
2. Comparison with LHC, CLAS

• Compared to LHC, Hall D has
• Similar (LHCb, BTev) or higher trigger rate
• Much smaller events
• Much smaller rate off detector
• Much smaller total trigger rejection
• Similar rate to tape
• Less cpu/evt needed in Level 3

6
2. Comparison with LHC, CLAS

• Compared to CLAS, Hall D has
• Much higher trigger rate
• 200 KHz vs 3 KHz
• Same size events
• Approximately the same number channels
• Much higher rate off detector
• 1 GB/s vs 25 MB/s
• Factor 10 Level 3 rejection
• CLAS has no Level 3
• Factor 4 higher rate to tape
• 100 MB/s vs 25 MB/s

7
Hall D
KTeV
KTev
CLAS
8
Atlas
BTev
Hall D
CMS
KTev, CDF, DO, BaBar, CLAS
9
3. Additional Considerations

• Can not interrupt ROC every event (200 KHz)
• Event blocking in front end cpus
• Timing and trigger distribution
• Note that CLAS has
• 25 crates
• 1 Trigger supervisor
• 1 Event Builder and 1 Event Recorder
• No Level 3 farm

10
Hall D DAQ Baseline Architecture
50-100 front-end crates
Gigabit switch 200 KHz
8 event builders
4 Gigabit switches
200 Level 3 Filter Nodes
4 event recorders
Network connection to silo 20 KHz
4 tape drives
11
3. Additional Considerations, cont

• Crates vs networked front end boards?
• If crates used, VME vs CPCI vs ?
• (RT)Linux vs VXWorks in front end cpus?
• Need low-latency interrupt in front end cpus?
• Location of electronics, crates?
• Grounding design?

12
4. DAQ Challenges

• All problems solved somewhere, many in CLAS
• But new to JLab/CODA
• Timing distribution
• Event blocking
• Many more front end crates
• Multiple event builders/recorders
• Large Level 3 farm
• Multiple, simultaneous DAQ systems (for
  commissioning)
• Need for fault tolerance
• Integration with control system
• How are we going to do it?
• See next talk

13
Backup slides
14
3. Comparison, cont
Event Size L1 Input Rate L1 output Rate L2 output Rate L3 output Rate
KTev 8 KB 100 KHz 800 MB/s 20 KHz 160 MB/s 2 KHz 7 MB/s
CDF 270 KB 50 KHz 13 GB/s 300Hz 80 MB/s 80 Hz 23 MB/s
D0 250 KB 10 KHz 2.5 GB/s 1 KHz 250 MB/s 70 Hz 13 MB/s
BaBar 33 KB (1200 L1) 2 KHz 2.4 GB/s None (65 MB/s) 100 Hz 4 MB/s
BTev 50-80 KB 800 GB/s 80 KHz 8 GB/s 4 KHz 200 MB/s
15
3. Comparison, cont
Event Size L1 Input Rate L1 output Rate L2 output Rate L3 output Rate
Atlas 1-2 MB 75 KHz 100 GB/s 3 KHz 5 GB/s 200 Hz 300 MB/s
CMS 1 MB 100 KHz 100 GB/s 100 Hz 100 MB/s
CLAS 6 KB 4 KHz 4KHz 25 MB/s 4KHz 25 MB/s
Hall D 5 KB 400 KHz 200 KHz 1 GB/s none 20 KHz 100 MB/s