From nch@roe.ac.uk Fri Oct 5 16:47:03 2012 Date: Fri, 5 Oct 2012 12:36:39 +0100 From: Nigel Hambly To: Eckhard Sutorius , Nicholas Cross , Mike Read , Ross Collins , Rob Blake , Mark Holliman , Clive Davenhall Cc: Lorenzo Rimoldini , Bob Mann , Jim Emerson , Keith Noddle , Norman Gray , Stephen Warren , Tom Shanks , Simon Dye , Stelios Voutsinas , Tom Kerr , Mike Irwin , Andy Lawrence at ROE , Nigel Metcalfe Subject: WFAU VDFS science archives project team meeting, 5 Oct 2012 Present: ETWS, RGM, NCH, NJC, RSC Apologies: JPE, AL, STV, NG, KTN, MSH, MAR, RPB DoNM: 10am Friday October 19th 2012 *** NB in the Villas meeting room *** MAR in the chair. Actions discharged this time: ----------------------------- ACTION: RSC to scope out the areas/situations where we'd like to test MonetDB i.e. where do the current bottlenecks lie? Discharged; detection catalogue ingest (pure ingest speed), quality bit flag updating, and the later CUs have all been benchmarked for the current system and results put on the TWiki. ACTION: NCH to contact VVV co-PIs to suggest expurgating the crud from VVV DR1 and making it world-public. Discharged; it has been agreed to try to expurgate the dodgy stuff from the existing VVV-DR1 and then make it world public. ACTION: RGM to chase up NG to find out what's happening with the UCD1+ Discharged; NG apparently working on an automated translation process that employs machine-learning techniques. Actions partly discharged but continuing: ----------------------------------------- ACTION: ETWS (with help from NJC as necessary) to add FeII 1.644 micron filter details into the WSA. Continues; all done apart from Vega-to-AB and zeropoint attributes (the later needs the latest data to be ingested). ACTION: NCH to check the GCS for pawsky-related issues, and liaise with SJW over the LAS results Continues; see Survey Release below (honestly) Actions carried forward from 7/09/12 meeting --------------------------------------------- ACTION: ETWS, with RPB, to develop an automated helper script to keep release DB files links updated. Continues; this one still has low priority. Specific points and new actions: -------------------------------- Project management: RGM noted that it looks like the WFAU grant will have to be "bridged" from April 1 2013 owing to the renewal review processes being stalled by the impending Programmatic Review. RGM is speaking to STFC this pm to discuss the details. WFCAM, VISTA and VST updates: The team noted the impressive prospectus that has been issued in support of the AO for take-over of UKIRT. Comments and issues arising from CASU minutes: No new minutes this time. Networking: ETWS noted -- The team at Chicago have setup with a UDR rsync link and transfers of SDSS BestDR9 are running now at 5MB/s. Routine transfers from CASU indicate that UKLight is holding up well: 8/20/35 MB/s for WFCAM/VISTA/VST data seem to be typical. WSA/VSA/OSA/GES Operations: Archive operations have been heavily impacted by unexplained problems with a RAID array over the last week or so. Random corruption of fits files as read off the arrays from certain machines has resulted in it being felt necessary to copy 30TB of data off of one unit onto another. ETWS noted -- Still running fverify tests on the data on disk38 and unfortunately files are still showing up as corrupted. -- Copying the data off disk38 and hopefully getting a usable copy on disk41 next week. Replacement of a RAID card in the disk38 NAS box has failed to completely solve the problem. Hardware and Systems: RSC reported: "Updated TWiki page for Monet DB investigations with the raw performance figures for the VVV for various database operations that are performed during regular curation, focused on the detection tables where scalability issues will most likely lie. This is entirely based on the DR1 experience to date, which does not include the changes made for DR2 to improve performance of detection tables by splitting them into monthly partitions. http://apache.roe.ac.uk/twiki/bin/view/WFAU/DatabaseScalabilityPerformanceAnaly sis The next stage in this process would be to obtain a full profile of performance of CU6 on the VVV which is currently ridiculously slow in taking more than 6 weeks to run. This will require a suitable gap in operations." ETWS reported: -- Tested the data storage at ECDF with transfer and CU2, 3 and 4 processing tests. The same tests were run also locally on disk41. Generally, tasks that write to the disks are slower than having the data here, reading is as well slow but not as significant. Results are compiled on the TWiki at WFAU Web > VDFSScienceArchives > SystemsManagement > EcdfMountedRaidTests e.g. updating FITS headers once the data are written runs ~2x slower while jpeg writing seems to be about 30% slower. Transfers run at 8 MB/s, which is unsustainably slow for combined WFCAM/VISTA/VST operations. Note that ECDF = Edinburgh Compute and Data Facilities, a central service provided by UoE. Survey Data Release: DR10 is still on-hold due to worries over LAS J sky subtraction quality since a ramp-up in use of the pipeline paw-sky algorithm since around late May 2011. Final decision rests with the survey head as to what to do about this in the quality control deprecations. NJC noted that we're also awaiting the final DR10 UDS stacks from Nottingham. Interesting gossip concerning VIKING, where a new survey head (= UKIDSS DXS survey head) has taken over. Much email communication to vsa-support has ensued to bring the new boss up to speed with all the release shenanigans. Software: RSC noted: "Reopened my investigations from earlier this year into performing curation tasks that only read data, such as EsoRelease, against read-only release databases, to avoid the need to take a release database off-line to run such tasks." NJC reported: "I spent much of the last few weeks testing problems with CU13, which ultimately came down to a strange error on /disk38 - files appearing corrupt on some servers but not others. At the same time I have added in the changes that ESO required for Phase 3 catalogues and have developed a new curation unit (cu29) to ingest and match survey team catalogues such as the ones that VMC have sent us. I will put some notes on the Twiki about this soon." Non-survey Data Release: Nothing of note this time. Astrogrid deployment, VO & Data Analysis services: RGM noted that the migration of Astrogrid hardware and services from Leicester to Edinburgh has taken on new urgency with the imminent removal of the HW at the UoL end. MSH has been working on this as fast as possible. Miscellaneous: The team bid a fond farewell to the Vista Hut, which has served us well as a meeting place since the demise of the venerable UKSTU Plate Library. ============================================================= Nigel Hambly Tel: +44-131-668-8234 Institute for Astronomy Fax: +44-131-668-8416 School of Physics and Astronomy University of Edinburgh Email: nch@roe.ac.uk Royal Observatory Blackford Hill Edinburgh EH9 3HJ The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336. ============================================================= -- Scanned by iCritical.