1. Modeling PANDA TDAQ system Jacek Otwinowski Krzysztof Korcyl Radoslaw Trebacz Jagiellonian University - Krakow

2.  Major Components and Functionality  All detector channels are self triggering entities  Frontend Electronics - hit detection and data pre- processing (hit time reconstruction, pattern reconstruction,... )  Time Distribution System - provides clock for hit timestamps  Concentrators/Buffers - buffering and on-the fly data flow manipulation  L1 Compute Nodes - mark hits which might belong to the same event (time slice)  L2 Feature Extraction Nodes – combine detector information to extract physical signatures (particle ID, momentum,...)  L3 Event Selection Nodes – event selection based on a complete reconstruction (particle IDs, event vertex, invariant masses,...)  Optical Links (connect detectors and buffers)  High Speed Network (1-10 Gbit/s bandwidth, switches) Panda TDAQ - System Overview

3. Panda TDAQ – Requirements Physics+Noise+Background+Pileup: 40-100 GB/s Data Storage: 100-200 MB/s  ~1000 reduction factor Event rates10 MHz1 MHz0.1 MHz Data reduction(100 GB/s)(10 GB/s)(1 GB/s) Latency 100  s 1 ms10 ms Nb. of Nodes1000 (10% of data) 1000 Example:

4. Physics Generator Panda DAQ – modeling concept Detector Noise Generator L1 - Processor Farm Data Storage Detector Buffer L0 L1 L2 L1 Results Buffer L2 Results Buffer L2 - Processor Farm L3 - Processor Farm Data (push) accept Data (pull) accept Data (pull) push SystemC - C++ open source language and simulation kernel for modeling and implementing electronic systems. http://www.systemc.org

5. Modeling PANDA TDAQ - Physics Generator Physics –physics type (int) –probability of physics type –physics amplitude (int) Physics Generator 6 Output Ports Detectors

6. Modeling PANDA TDAQ – Detector (physics input case) Noise Generator Detector OutPort InPort Detector –detector efficiency F(physics type) –signal amplitude F(#detector) - bytes –noise probability F(#detector) –noise amplitude F(#detector) - bytes hit timestamp is assigned

7. Modeling PANDA TDAQ – Detector (noise case)

8. Modeling PANDA TDAQ – Detector Buffer (detector input case) Detector Buffer L2 L1 L0 3 Ouput Ports 4 Input Ports 3 Stage Buffers Detector Buffer –trigger latency F(trigger level) –Stage Buffer map (timestamp, *message) Reading Process Writing Process

9. Modeling PANDA TDAQ – Processor Farm1 (detector buffer input case) Writing Process Reading Process Processor Farm1 7 Input Ports 6 Input BuffersOutput Buffers 7 Output Ports Processor Farm1 –trigger latency F(trigger level) –Buffers map (timestamp, *message) event number is assigned

10. Modeling PANDA TDAQ – Detector Buffer (L1 farm input case)

11. Modeling PANDA TDAQ - status Skeleton coding startedSkeleton coding started simple model of the physics generator is ready detector and noise generator merged in a single detector model models of detector buffers and L1 processor implemented communication between modules (ports, channels, read and write interfaces) implemented Plan to close the event loop (from physics generator to storage) by the end of year – this will open a way to implement more realistic models of contributing components and also use realistic parameters for critical places in the architecture (latencies, throughput, processing times…)