Presentation is loading. Please wait.

Presentation is loading. Please wait.

CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 A Control Software for the ALICE High Level Trigger Timm.

Similar presentations


Presentation on theme: "CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 A Control Software for the ALICE High Level Trigger Timm."— Presentation transcript:

1 CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 A Control Software for the ALICE High Level Trigger Timm M. Steinbeck Computer Science/Computer Engineering Kirchhoff Institute of Physics Ruprecht-Karls-University Heidelberg, Germany

2 CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 Outline – Overview ALICE / High Level Trigger (HLT) – Overview Control System / Requirements – Architecture ● Configuration ● Program State Handling ● Program Interface ● Master/Slave Operation – Sample Configuration – Summary & Conclusion

3 CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 Overview - ALICE ALICE High Level Trigger – ALICE: A Large Ion Collider Experiment – Heavy-Ion Mode: ● Up to 15000 particles in event ● Max. event size >70 MB ● Max. input rate from TPC: 200 Hz ● Input data stream:  25 GB/s ● Output data stream:  1.2GB/s – Proton-Proton Mode: ● Max. input rate from TPC: 1kHz ● Event size  3 MB See also Talk by M. Richter and first talk by T. M. Steinbeck

4 CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 Overview - HLT ALICE High Level Trigger – Processing in several steps from raw detector data to full event reconstruction – Large PC cluster ● Initially  400-500 nodes ● Arranged in hierarchy levels (HL) that match detector layout and analysis steps – Exact processing sequence and hierarchy not known – Framework consisting of multiple independant, communicating components used to transport data – Software needed to manage components and their interaction

5 CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 Outline – Overview ALICE / High Level Trigger (HLT) – Overview Control System / Requirements – Architecture ● Configuration ● Program State Handling ● Program Interface ● Master/Slave Operation – Sample Configuration – Summary & Conclusion

6 CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 HLT Control System Purpose of the HLT Control System: Management of HLT Processes – Supervise processes – React to state changes ● E.g. errors, unexpected termination – Manage system startup ● Connect components only when they are ready ● Connect bridges between nodes when nodes are ready – Orchestration of HLT system

7 CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 Requirements – Flexible ● HLT Configurations not yet defined ● Usable for other programs as well – Hierarchical ● >2000 processes not manageable by single supervisor instance ● Eases configurations – No single point of failure ● PC cluster nodes are unreliable ● No need for highly reliable, expensive, equipment – Should be able to run on the cluster nodes themselves, along the analysis processes.

8 CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 Outline – Overview ALICE / High Level Trigger (HLT) – Overview Control System / Requirements – Architecture ● Configuration ● Program State Handling ● Program Interface ● Master/Slave Operation – Sample Configuration – Summary & Conclusion

9 CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 Core TaskManager Architecture Configurati on Engine XML Configuratio n File Interface Library Controlled Programs Program State & Action Engine Embedded Python Interpreter TaskManager Program Interface Engine

10 CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 Core TaskManager Architecture Configurati on Engine XML Configurat ion File Interfac e Library Controlled Programs Program State & Action Engine Embedded Python Interpreter TaskManager – TaskManager configuration is stored in XML files – "State machine" logic is Python code, read from configuration file and executed by embedded Python interpreter – Communication with external programs is done via external interface library, supplied with programs, specified in configuration file Program Interface Engine

11 CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 Outline – Overview ALICE / High Level Trigger (HLT) – Overview Control System / Requirements – Architecture ● Configuration ● Program State Handling ● Program Interface ● Master/Slave Operation – Sample Configuration – Summary & Conclusion

12 CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 Configuration Files – Configuration is stored in XML files – Configuration Engine reads in configuration – TaskManager Core queries configuration items from Configuration Engine – Configuration files store ● Programs to be started and controlled ● Python code for "state machines“ of controlled programs and global ones ● Interface libraries to be used for communication with programs ● Master/Slave TaskManager configuration items Core Configurati on Engine XML Configurat ion File Interfac e Library Controlled Programs Program State & Action Engine Embedded Python Interpreter TaskManager Program Interface Engine

13 CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 Configuration Files – For each controlled program the configuration can describe: ● The command line to be executed ● The address for communication with the program (passed to the interface library) ● Resources used by the program (for cleanup upon termination) ● Different situations in which to execute Python code include – State changes – Termination – Configuration change Core Configurati on Engine XML Configurat ion File Interfac e Library Controlled Programs Program State & Action Engine Embedded Python Interpreter TaskManager Program Interface Engine

14 CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 Configuration Files – Global items in a configuration file include: ● The interface library to be used for communication with controlled programs ● Global state change action Python code ● Slave control specification (only slave TM) ● Master TM configuration is handled via normal controlled program specification Core Configurati on Engine XML Configurat ion File Interfac e Library Controlled Programs Program State & Action Engine Embedded Python Interpreter TaskManager Program Interface Engine

15 CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 Outline – Overview ALICE / High Level Trigger (HLT) – Overview Control System / Requirements – Architecture ● Configuration ● Program State Handling ● Program Interface ● Master/Slave Operation – Sample Configuration – Summary & Conclusion

16 CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 Program State Handling – „State Machine“ logic contained in Actions, Python code in configuration file – TaskManager has embedded Python interpreter – When conditions are met interpreter executes specified code – More powerful and flexible than pure state machine – Python code is specified for different occasions, e.g.: ● Program state change ● Program termination ● Configuration changes – Set of special functions provide interface from Python code to TaskManager and controlled programs Core Configurati on Engine XML Configurat ion File Interfac e Library Controlled Programs Program State & Action Engine Embedded Python Interpreter TaskManager Program Interface Engine

17 CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 Program State Handling – Python TaskManager Interface contains functions for multiple purposes ● Query state and status data of controlled programs ● Send commands to controlled programs ● Start programs ● Terminate (forcibly) programs ● Exchange information between master & slave TaskManagers Core Configurati on Engine XML Configurat ion File Interfac e Library Controlled Programs Program State & Action Engine Embedded Python Interpreter TaskManager Program Interface Engine

18 CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 Outline – Overview ALICE / High Level Trigger (HLT) – Overview Control System / Requirements – Architecture ● Configuration ● Program State Handling ● Program Interface ● Master/Slave Operation – Sample Configuration – Summary & Conclusion

19 CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 Program Interface – Programs use different methods to present status to the outside (including the TaskManager) –  Communication between TaskManager and programs is handled by interface libraries – Interface libraries ● are external to the TaskManager ● belong to a set of programs ● are loaded at runtime ● know how to communicate with their programs ● implement specific, defined set of functions – TaskManager can call specified functions ● Consistent interface API for all programs Core Configurati on Engine XML Configurat ion File Interfac e Library Controlled Programs Program State & Action Engine Embedded Python Interpreter TaskManager Program Interface Engine

20 CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 Program Interface – Program interface includes functions to ● initialize the interface library ● deinitialize the interface library ● query a program's state ● query additional status data for a program ● send a command to a program ● wait for interrupts (or look-at-me-s) from programs Core Configurati on Engine XML Configurat ion File Interfac e Library Controlled Programs Program State & Action Engine Embedded Python Interpreter TaskManager Program Interface Engine

21 CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 Outline – Overview ALICE / High Level Trigger (HLT) – Overview Control System / Requirements – Architecture ● Configuration ● Program State Handling ● Program Interface ● Master/Slave Operation – Sample Configuration – Summary & Conclusion

22 CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 Master Slave Configuration – Slave TaskManager uses additional internal object to receive commands from master and provide master with status data – Master uses interface library to communicate with slaves – Master/Slave communication is simple extension to standard communication w. controlled programs Core Configurati on Engine XML Configurat ion File Interfac e Library Program State & Action Engine Embedded Python Interpreter TaskManager Program Interface Engine Core Configurat ion Engine XML Configurat ion File Interfac e Library Controlled Programs Program State & Action Engine Embedded Python Interpreter TaskManager Program Interface Engine Master/Sla ve Interface

23 CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 Outline – Overview ALICE / High Level Trigger (HLT) – Overview Control System / Requirements – Architecture ● Configuration ● Program State Handling ● Program Interface ● Master/Slave Operation – Sample Configuration – Summary & Conclusion

24 CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 Sample Configuration – Configuration planned for HLT use in ALICE TPC Testbeam – 2 optical readout links from detector via DAQ splitter module – 1 optical output link going to DAQ – 3 PCs, one per optical link, several software components on each PC HLT PC 1 RORC Publisher Subscriber BridgeHead HLT PC 3 Publisher BridgeHead Publisher BridgeHead Event Merger Storage Writer HLT-Out Subscriber EventRate Subscriber HDD Optical Link From DAQ/Detector HLT PC 2 RORC Publisher Subscriber BridgeHead Storage Writer HDD Optical Link To DAQ Optical Link From DAQ/Detector

25 CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 Sample Configuration Master TaskMana ger Slave TaskMana ger 1 Slave TaskMana ger 2 RORC Publish er Storage Writer Subscribe r BridgeHea d Publisher Bridge Head Publisher Bridge Head HLT-Out Subscrib er Event- Rate Subscrib er – Status & Command flow is black HLT PC 2 HLT PC 3 Event Merger Slave TaskMana ger 0 RORC Publish er Storage Writer HLT PC 1 Subscribe r BridgeHea d – Data flow is yellow

26 CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 Outline – Overview ALICE / High Level Trigger (HLT) – Overview Control System / Requirements – Architecture ● Configuration ● Program State Handling ● Program Interface ● Master/Slave Operation – Sample Configuration – Summary & Conclusion

27 CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 Summary & Conclusion – TaskManager provides flexible and hierarchical program control – Powerful program state handling through embedded Python – Adaptable to many programs through interface libraries – Machine- and human-readable/writeable XML configuration files – Avoids single-points-of-failure – Used for ALICE High Level Trigger Data Challenges – Used for HLT control in ALICE TPC testbeam – Designed to be adaptable to other problems, in- and outside of Alice HLT


Download ppt "CHEP04 - Interlaken - Sep. 27th - Oct. 1st 2004T. M. Steinbeck for the Alice Collaboration1/27 A Control Software for the ALICE High Level Trigger Timm."

Similar presentations


Ads by Google