90 likes | 171 Views
DAQ thoughts about upgrade. 11/07/2012 Barthelemy.Von.Haller@cern.ch Sylvain.Chapeland@cern.ch. What we would like to see in a future framework (1). Unique database for online and offline configs and calib , include data valid for current run (for DA and DQM per instance)
E N D
DAQ thoughts about upgrade 11/07/2012 Barthelemy.Von.Haller@cern.ch Sylvain.Chapeland@cern.ch
What we would like to see in a future framework (1) • Unique database for online and offline configs and calib, include data valid for current run (for DA and DQM per instance) • Data traceability : calibration -> ocdb -> application (keep track of calibration and config used for run N) • Unify data format that could be used throughout systems (from LDC). If two format (like raw and ROOT, then transparent conversion from one to the other) • Review our policy on dropping incomplete events • Single lib/interface to read/write data (local, castor, tags, indexes)
What we would like to see in a future framework (2) • Independent modules of reasonnable size • Make use of new available online CPU (for reco or analysis) • Dynamic adaptation of algorithms to available online resources (e.g. online reco on event building cpus for a fraction of events) • DQM-QA merge : unique system usable online and offline for all QA related tasks. • DQM : Reco online of events and storage of results for later. Completion of QA offline, but first X % online. • Multithreading
Common Procedures (1) • Not necessarily common hardware or facilities ! • E.g. same build system know-how but different instances • Release numbering • Build system & Packaging • Tools • e.g. cmake, svn, rpm, yum, ticket system • Dependencies handling (aliroot version to use in prod/test) • Benchmark (ref sys and tools) • Mock modules
Common Procedures (2) • Code access policy (open-source ? Commit rights ?) • Languages (C++, standard libraries STL/Boost, ROOT) • OS (Production farm and desktops ; SLC, Ubuntu, Mac OS, Windows, mobile) • HW (x86_64, GPUs, virtualisation) • Storage (cloud ?) • Core language features (e.g. log / printf / cout) • Coding guidelines and standards
Current daq tools or features that could be used/replaced by a common framework • Data validity checks, e.g. what we have in the readout now • Infologger • Logbook (common source of information for online and offline) • FXS (Move data around at runtime between system) • Publish-Subscribe and notification mechanism (now DIM) • Process control • AMORE (online DQM allowing parallel execution of detector code, collection of results and visualization through custom, generic and web user interfaces)
New DAQ • Replay and inject events at any level (testing, benchmarking, mock) • Started a prototype of a new DAQ system based on modular IO boxes • Few types of modules: event producer, consumer, filter • C++ base classes provide common needs: FIFOs, thread pools, process control • Inheritance + virtual methods to implement custom behavior of each module • Flexible description of modules interconnects to instantiate different DAQ systems
Answers to Matthias’ questions • Max data recording : 4GB (4x10Gbits fibers to CASTOR) • Limiting factor : bandwidth to CASTOR (4 vs >8 GB locally) • Easy scale up of the system (more LDCs and GDCs) and hardware evolution • Other « DAQ » group services : ACT, ECS, FXS, Logbook, DQM, DA, ACR stations, InfoLogger, DIP
Conclusion • We are looking for topics of reflexion and debate , discussions and definition of architecture and common processes, but not for direct technical solution or implementation. • The Panel should also define the processes of R&D for what will follow the Panel’s work and provide requirements.