The new portal is available for testing and validation

https://accounting-next.egi.eu/

The views we are mostly interested in are 'WLCG' view and 'Reports'

Please provide your feedback and describe detected problems in the sections below

Detected problems

  • Double counting of the Desy statistics because of Desy belonging to several per-VO Germain federations (Fixed)
  • "Wall Time (h)" is lower than "CPU Time (h)", "Normalised Wall (h)" is also lower than "Normalised CPU (h)". Problem in number of cores? Something else? (Fixed)
    The naming schema on the WLCG tab should probably be re-used on the landing page portal.png
  • The "CPU Efficiency" table sometimes show 1% differences when compared with the CSV/JSON available on the same page. Rounding problem? Better rounding? Or perhaps one more significant digit needed?

Suggestions

  • Everything relevant for WLCG should be available under WLCG view, for example aggregation by country, or VO manager view for WLCG VOs (Julia)

  • I would suggest that the very first plot on the page (just below the table) can be shown in a different form: line or stacked bar. I would suggest stacked bar is a default one. In some cases accumulative plot also makes sense. A possibility to get a plot in a different form should be provided in the upper part menu (Julia)

  • Would it be possible to select number of series shown separately in the plot , while all others are summed up. For example in this view in the plot in the bottom of the page I get 6 results, 5 top sites and all the others grouped. If I would like to have top 15 sites and all other grouped what should I do? I suggest you add a possibility to define number of time series shown separately on the plot for every plot, where number of items is higher than let's say 5.

  • From the input from Simone (below) as well as from Ian Bird and some site admins: it would be very useful to have a bar plot where consumption is shown with bars and at the same plot we have a pledge shown with the line. Since it is not relevant for all kinds of selection, it can be shown under a dedicated selection option under the WLCG view, something like "consumption vs pledges". (Julia)

What experiments need from the portal for preparation of the scrutiny reports

ATLAS, Input from Simone

We take all pledges from Rebus. This has nothing to do with the accounting portal, but a reminder we do need pledges from Rebus. The total Used for CPUs refers to WallTime. The efficiency is CPU/Wall and therefore in general we need both CPU and Wall time.

CERN CPU: a can of worms. It is a combination of T0 resources + local batch + machines for central services. What we need here from WLCG accounting is the WallTime and CPU time of T0 resources (they are in a dedicated ATLAS LSF instance). That will account for those resources when they run T0 jobs and also Grid jobs.

T1 and T2 CPUs: we need CPU time and WallTime consumed by ATLAS. It needs to be separate for T1s and T2s of course. Not critical, but used some time is the same information at the level of the site and country. We use this for debugging purposes and to answer special questions of the CRSG. So, we do not strictly use it to write the report but we use the information often. We need this information with the daily granularity (as sometimes we need to report e.g. 3 months from 15 of february to 15 of may). It is very useful to have monthly summaries as well, which is what we mostly use, for example to generate the plot in attachment with the monthly consumption with respect of pledge.

The HLT CPUs are monitored only in the dashboard and we use that one. HPCs are monitored only in the dashboard and also there this is what we use.

We take all the “Total Used” disk information from SRM, no need of this from the WLCG accounting. We take all the tape information from the dashboard, no need to do anything in WLCG accounting also there.

A special mention to HS06. We need a “as reliable as possible” average HS06/core slot at the level of the site (we do not use machine level granularity). This is useful when you want to translate the number of running jobs you observe with what you expect from the pledge (which is in HS06). We assume an average 11HS06/core, which is the average we get from rebus which is I guess about right as it is the global WLCG average and mistakes at the level of the single site average out. But would be nice to have the number as reliable as possible. Again, not strictly used in the report, but user rather often to answer questions.

Table with data for the scrutiny report

Plot would be useful to get on the portal

CMS, Input from Pepe

CMS Presentation

ALICE input

LHCb input

-- JuliaAndreeva

Topic attachments
I Attachment History Action Size Date Who Comment
PDFpdf 20160622_CMS_ResourceUtilization.pdf r1 manage 363.5 K 2016-06-29 - 16:16 JuliaAndreeva  
PNGpng Screen_Shot_2016-06-29_at_1.00.33_PM.png r1 manage 221.3 K 2016-06-29 - 13:00 JuliaAndreeva  
PNGpng Screen_Shot_2016-06-29_at_1.01.58_PM.png r1 manage 168.1 K 2016-06-29 - 13:02 JuliaAndreeva  
PNGpng portal.png r1 manage 212.6 K 2016-07-06 - 13:33 MiguelSantos  
Edit | Attach | Watch | Print version | History: r14 | r9 < r8 < r7 < r6 | Backlinks | Raw View | Raw edit | More topic actions...
Topic revision: r7 - 2016-07-18 - JohnGordon
 
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    LCG All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright &© 2008-2019 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback