130 likes | 253 Views
FAX Deployment, Service and Storage Integration . Wei Yang. Overview. FAX Components and Services Redirector, LFC and monitoring Infrastructure Sites deployment Status Use cases Panda Cloud On-going Integration with Storage systems R&D Activities. Infrastructure and Services.
E N D
FAX Deployment, Service and Storage Integration Wei Yang US ATLAS Distributed Facility Meeting University of California Santa Cruz
Overview FAX Components and Services • Redirector, LFC and monitoring Infrastructure • Sites deployment Status Use cases • Panda • Cloud On-going Integration with Storage systems R&D Activities US ATLAS Distributed Facility Meeting University of California Santa Cruz
Infrastructure and Services • A Network/Tree of Redirectors • Allow a user to start from anyway and reach everywhere • Multiple levels of redirectors • Top level: EU & BNL • Country level: DE, FR, RU, UK • Regional level: US central (hosted by UC) • Site level: UC, SLAC • Read-only LFC services • Hosted by BNL (for US sites) and CERN (for EU sites) • Monitoring Data Collectors • Collect and send monitoring data to ATLAS dash board • Site specific/unique file for validation US ATLAS Distributed Facility Meeting University of California Santa Cruz
http://ivukotic.web.cern.ch/ivukotic/FAX/index.asp BNL and EU redirectors are peers at top level due to network latency US ATLAS Distributed Facility Meeting University of California Santa Cruz
The Monitoring Services • Availability Dashboard • Current running at UC, will be migrated to ATLAS SSB • Detail Monitoring Collector • A.K.A UCSD collector, collect info on every read • Aggregated info file level access info • Send to ATLAS monitoring dashboard via ActiveMQ • Summary Monitoring Collector • Based on MonaLisa, aggregated at data server level • Info used to compare with detail info and debugging • ATLAS Monitoring Dashboard for FAX • Integrate with AGIS US ATLAS Distributed Facility Meeting University of California Santa Cruz
https://uct3-xrdp.uchicago.edu:8443/rsv/ US ATLAS Distributed Facility Meeting University of California Santa Cruz
From Julia Andreeva FAX Dashboard and ML FAX repository comparison FAX Dashboard now includes EOS, which dominates over all other transfer/access This plot is showing overall traffic rate over last 12 hours group by source , excluding CERN (EOS) Aggregated xrootd traffic rate over last 12 hours according to FAX ML repository, excluding MWT2_UC and SLAC which are missing in Dashboard In general is a good agreement, as well as going site by site. Big progress over last couple of weeks
http://dashb-atlas-xrootd-transfers.cern.ch/ui US ATLAS Distributed Facility Meeting University of California Santa Cruz
Site Deployment • 8 sites in the US (all sites) • 4 sites in the UK • 3 sites in DE • 2 sites in RU • 1 site in Prague, CZ • working with IT cloud https://twiki.cern.ch/twiki/bin/view/Atlas/FaxSiteCertification US ATLAS Distributed Facility Meeting University of California Santa Cruz
Use Cases • Interactive Access from Desktop/Laptop • Xrdcp or ROOT/ProofLite • From Panda Jobs • Prun: supply a list of files in global name • Panda pilot support • Phase I: replace missing files using FAX • See Paul’s talk. Expanding testtomore Tier 2 sites • Phase 2: use site cost matrix for job scheduling • Phase 3: beyond, a lot more opportunities … See Torre’s talk • By the Cloud • FAX is a nature choice for jobs in the Cloud to consume data • Inbound data traffic is free/low cost (outbound is expensive) • No need for long term storage in the Cloud US ATLAS Distributed Facility Meeting University of California Santa Cruz
Storage System Integration • Have solutions for almost all ATLAS systems • Basic idea: • A dedicated xrootd machine to help the site joining FAX either as a helper, refer client to the site storage or a proxy, fetch data from site storage on client’s behave • Translate global file name to site storage file name • Support POSIX (NFS, Lustre, GPFS, etc.), Xrootd (including EOS), dCache, DPM • Working on Castor (RAL) • Support • tWiki and mailing list • https://twiki.cern.ch/twiki/bin/viewauth/Atlas/AtlasXrootdSystems • atlas-adc-federated-xrootd@cern.ch • Bi-weekly Vidyo meeting on deployment issues • Experts in the US for general Xrootd and dCache support • UK/DPM team support DPM integration to FAX • Some sites are creative and self support (EOS) • Cloud level support: e.g. DE and UK clouds US ATLAS Distributed Facility Meeting University of California Santa Cruz
R&D • Driven by feature request/Operation feedback • Deployment and Operation are the focus • But some level of R&D is still needed for a while • Have experts in many R&D area in US and EU • R&D provides • New functions/features, e.g. f-stream • bug fixes • New models for site and ADC specific needs US ATLAS Distributed Facility Meeting University of California Santa Cruz
Federated Xrootd deployment timeline …more dCachedev …new monitoring stream & integration issues As always, the docs could Be better From Rob Gardner