Online Software Timing Tests for Large Scale Tests 2004:
-------------------------------------------------------
-------------------------------------------------------
Result files are under
/afs/cern.ch/user/a/atlonl/public/LST04/results/processed
and can be viewed with gv; The last plots in the file contain the summary
information on the transitions for all partitions.
known problem:The last page 'successful measurements per run versus partition
size'
is not correct.
all: no OBK, no DDC
res400to1000ctrleli.ps
----------------------
run during the day; Wednesday 10. March; max 146 hosts; 5 cycles per partition
first test series; large fluctuation and increased values for boot time from 6
50 ctrls on; timeout parameter was not large enaugh and increased in local
play_daq
res10to100ctrldoris.ps
----------------------
run during the day; controllers only;
8.March; max 111 hosts
O.K. ~ 5 cycles per partition
res50to1000ctrldoris.ps
-----------------------
night tests ;12. March ; controllers only;
max. 146 hosts; ORB parameters
set; O.K.
timing_results_50to1000_mon.ps
------------------------------
night test; Monday 15.March; controllers, one
monitoring task per local controller
to be startet at boot time and one monitoring sampler per detector controller
to be
started user defined (so not yet
started); mo max 216 hosts; ORB
parameters set
Close time: quasi constant from 200 ctrls on until 800 ctrls, then steep
increase!!??
shutdown timeout of 30s at 700 ctrls reached
timing_test_mon_50to1000Tuesday.ps
----------------------------------
night test Monday Tuesday 16. March;
ORB parameters set
second night mon partitions; complete cleanup and start of agents before
starting
controllers, one monitoring task per
local controller to be startet at boot time
and one monitoring sampler per detector controller to be started user defined
(so
not yet started); ORB parameters set
Close time: quasi constant from 200 ctrls on until 800 ctrls, then steep
increase!!??
shutdown timeout of 30s at 700 ctrls reached
timing_result_mon_50-1000Thursday.ps
------------------------------------
night test Thursday 18th; max 216 hosts;
ORB parameters set;
controllers, one monitoring task per
local controller to be startet at boot time
and one monitoring sampler per detector controller to be started user defined
(so
not yet started); ORB parameters set
improved configuration access after patch 4
Close time: quasi constant from 200 ctrls on until 800 ctrls, then steep
increase!!??
shutdown timeout of 30s at 700 ctrls reached (need a 'dsa_stop-supervisor,
meanwhile available under /afs...atlonl/eli)
timing_result_rdb_friday_doris.ps
---------------------------------
night test Friday 19th; controllers only; max 145 hosts; ORB parameters set
ORB parameters set
shutdown values largely scattered; network problem? to be investigated; a
number of test cycles not finished; interesting scatter plots;
timing_results_rdb_saturday_doris.ps
------------------------------------
night test Saturday 20th ; controllers only;
max 145 hosts; ORB parameters set
ORB parameters set
shutdown values largely scattered; network problem? to be investigated; a
number of test cycles not finished; interesting scatter plots;
timing_results_Tuesday_mon_GB_corbaloc.ps
-----------------------------------------
monitoring partitions run using corbaloc for ipc, on the
GB farm (only) on 120 hosts ; good results
scattering redulced very significantly as compared to previous test;
no lost test runs for the upper half of the partitions (the large
configurations) but 11 missing for the smaller partitions
for 20-50:
1021
controllers, 1000 mon samplers, 20 mon. tasks
boot: ~95 s, warm start 2 s, close 5 s,