1 / 17

Update on resource requests

Update on resource requests. Concezio Bozzi INFN Ferrara LHCb NCB , June 16 th 2014. C-RRB. The CRSG approved our model to compute requests for 2015 and 2016 ( LHCb-PUB-2014-014 ) However, there was a debate regarding the expected LHC livetimes in 2015 and 2016

elan
Download Presentation

Update on resource requests

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Update on resource requests Concezio Bozzi INFN Ferrara LHCb NCB, June 16th2014

  2. C-RRB • The CRSG approved our model to compute requests for 2015 and 2016 (LHCb-PUB-2014-014) • However, there was a debate regarding the expected LHC livetimes in 2015 and 2016 • the other experiments used running times (3 and 5Ms in 2015 and 2016) estimated almost one year ago, which eventually made their way into the document on the computing model update of the LHC experiments (CERN-LHCC-2014-014; LCG-TDR-002) • The LHCb requests were based on the current LHC schedule, and assumed 4 and 6Ms running times • The C-RSG requested that we use the same running times as the other experiments • These requests were approved by the C-RRB • LHCb and others remarked that this running scenario might be pessimistic, so the corresponding requests in the C-RSG report are NOT on the conservative side. • Discussion is continuing with LHCC referees (last meeting: June 3rd)

  3. 2015 resources M. Cattaneo@ LHCC referees, June 3rd • Concern about length of 2015 run • All experiment computing resources requests based on 3M seconds of LHC live time • And 5M seconds in 2016 • Numbers based on 2010, 2011 experience • Surely more efficient commissioning next year? • We know there are more days of physics scheduled in 2015 than in 2010 • We know that LHCb will saturate trigger rate from the beginning • Too late to change 2015 requests, but can we be more realistic (or less pessimistic) for 2016? • Ideally, would like an official statement on the 2016, 2017 and 2018 schedules from the CERN management • LHCC can help?

  4. Updated and approved requests

  5. Updated and original requests

  6. Updated and original requests 2015 original vs. updated: 10-15% diff at T0+T1 (T1 disk unchanged), 15%-20% at T2 2016 original vs. updated: 15% at T0+T1, 10% or less at T2

  7. Updated and original requests 2015 original vs. updated: 10-15% diff at T0+T1 (T1 disk unchanged), 15%-20% at T2 2016 original vs. updated: 15% at T0+T1, 15% less T2 CPU, 5% less T2 disk

  8. Updated and original requests 2015 original vs. updated: 10-15% diff at T0+T1 (T1 disk unchanged), 15%-20% at T2 2016 original vs. updated: 15% at T0+T1, 15% less T2 CPU, 5% less T2 disk

  9. Updated and original requests • Cost of current implementation of data preservation: 10% of tape requests • Original requests: 3.7PB in 2015, 7.3PB in 2016 • Updated requests: 3.2PB in 2015, 6.0PB in 2016

  10. Comparison with “flat budget” • Definition of flat budget: same money will buy • 20% more CPUs • 15% more disk • 25% more tape

  11. Comments and recommendations from the C-RSG

  12. Specific comments to LHCb • The disk requirements are driven by the split between the full DST and the mDST, as well as the number of copies of each dataset retained on disk. LHCb is asked to study the latter carefully in the coming months to determine whether 4 replicas of the current stripping dataset are truly required. • Significant tape space has been allocated for data preservation. […] LHCb is strongly encouraged to study the need for the second copy of the derived dataset. Some trade-off between cost, data integrity, and the time needed to reproduce the derived data — including the preservation of software and hardware environments — might be possible. This issue should be explored in more detail during the autumn scrutiny period.

  13. Data popularity: an example from ATLAS ATLAS says these unused data are Recently created “production” data (Production tasks can take more than three months) Data for which there is no explicit lifetime or “move to tape” policy Small-file data The “unused data issue” is understood and is being addressed.

  14. Next steps • Next scrutiny round (october 2014 – deadline for documents in August): • quantitative answers for • data popularity (waiting for metrics from C-RSG) • data preservation • Scrutiny of 2016 requests • Computing resources: • Tier1 countries: plan for big jump in tape • x1.7 in 2015, x3 in 2016 • All countries: prepare for 50% increase of CPU and disk in 2016 • Try to plan according to the original requests

  15. Other news • 0.5PB more disk pledged for 2014 at T2-D centers  1.26PB in total now at T2-D • Already in 2014, Russia is pledging a Tier1 (Kurchatov Institute for LHCb) • 10.8kHS06 CPU • 0.96PB disk • 1.14PB tape • CPU used since last fall for MC production • Looking forward to deploying storage resources • 0.17PB disk currently installed and commissioned

  16. backup

More Related