Hi all,
A happy new year to all of you. Apart from all our regular accounting and billing, there are also a few nice graphs to extract from our accounting data. For the Nikhef part, a couple of nice figures for your enjoyment:
In 2008, we processed 1441182 jobs, and on average 660 cores were in use averaged over the year. This is the equivalent of a little over 1 million days of work on a 1Ghz Pentium-III system, so if we had used pizza0-class boxes to do this work, we would have needed 1415 of them, or 70 racks full. The average CPU/wall efficiency was 86.3%
December was the busiest month, with 221200 jobs, of which quite a sizeable faction (actually most) were submitted during the New-Year break. So, we can happily go on holidays, it's good for business :-)
Also, 32 unique groups used our facility, from 25 different user communities (of which only 4 are dedicated to infrastructure testing).
Many job submissions are also automated, or people like working over the week-end. Although working days see a slightly higher submission rate, it's not more than ~ 20% over the week-end submission rate. (see the graph below). And Wednesday is the most popular day to start working on the grid.
In general, we see more submissions over the course of 2008, also prompted of course by the number of cores available. If you look at job submissions per core, it is rather constant (which means that in forst order all nodes we will buy and commissions will just be used. There is no slack in demand :-). Nice detail: if you reduce the capacity, like what happened to us is May/June 2008, the jobs keep coming, and thus job the submission count per core goes up.
Enjoy the database!
Cheers, DavidG.
