Sn

1
VOCE Operational Status
Enabling Grids for E-sciencE
VOCE Operational Status, April June 2006 Jirí
Chudoba, Jaroslava Schovancová, CESNET and
Institute of Physics, Academy of Sciences, Czech
Republic
Site status (June 22, 2006).
We tested stability and availability of VOCE
Resources Testing during spring 2006. This period
was splitted into two Seasons. The Season I
lasted from April 22nd to May 17th 2006 and the
Season II lasts since June 4th 2006 till end of
June. During the Season I we used the LCG 2.4 RB,
during the Season II we use the gLite 3.0
RB. This VOCE Resources Testing covers 13 CEs
listed in the table above, while sites
egee.srce.hr and eszakigrid66.inf.elte.hu are
under cover since the May 11th and site
ce.ui.savba.sk is covered since the June 5th. The
other sites are covered since the April 22nd
2006.
The figure above shows availability of all tested
CEs during both Seasons. The black square
denotes that CE was on dynamically obtained list
of CEs, the red circle symbolizes otherwise.
Twice a day we submit 50 very short jobs (less
than 1 sec of CPU time) to each CE dynamically
obtained glite-job-list-match --vo voce
test.jdl during the normal CE operation activity
(this is not a challenge!). Submission to single
CEs is time-shifted, we start submission to
specific CE, then wait 5 mins, and after these 5
mins we start submission to another CE. Let's
assign T0 to the time when the 50th job was
submitted. According to the paragraph above, the
T0 timetick differs from CE to CE. We check the
job status every 10 minutes until number of jobs
with states "Done (successfully)" or "Aborted"
reaches the number of submitted jobs, or until
time exceeds the maximal value Tmax. We cancel
all the affected jobs after time time Tmax.
In one case we tested reliability with 1 hour
jobs (20 jobs per CE), the results are shown on
the figure on right.
Here we show results for 3 time values, T110
mins, T21 hr and T36 hrs. At time Ti we
compare reliabilities of all CEs. When the
site undergoes the SFT, user gets
information whether the site passed or
failed. However, from our testing user
obtains more detailed information, such as
percantage rate of successful jobs and its time
evolution. Moreover, our testing shows the
resources as seen by a common user, while SFT
tests often perform under special
conditions. The day-averaged CE reliability time
evolution is shown on figures around. The green
bar represents percentage rate of jobs "Done
(successfully)" after T1, The blue bar
illustrates percentage rate of jobs "Done
(successfully)" after T2 and the red bar
symbolizes percentage rate of jobs with status
"Done (successfully)" after T3. The red point
denotes the percentage rate of "Running" jobs
after T3. The yellow bar represents days when no
job was submitted. The empty space (bar or red
point is missing) indicates that some of the
submitted jobs were aborted by the CE itself.
Poster presented at GRID'2006 International
Conference, June 26 - 30, 2006 in Dubna, Russia.
http//egee.cesnet.cz/