Recently I have noticed that, after an update, the values for the client colomns changed:

for Progress from 18 % to 54 % and
for To Completion from approx 600 h to 106 hours.

The progress jump from 18 to 54 would indicate that the total model time (45 years) had been reduced to 15 years.

This jump would have been plausible if the model was not running well and with the update this total time had been reduced from 3 Phases to just one.

But, running on to 66%, the computation was interupted for a CPU Benchmark and upon resumming the Progress value jumped back from 66 to 22%. (and completion time reflected 45 year also)

Therefore, I must conclude that in certain conditions, the software must revert to a wrong total time and thus corrupt the values in question.

Would like to know if other users experiance the same behaviour and if a developper can find anything wrong in the software, causing this minor, but confusing inconviniance.

B.t.w. running hadsm3 4.04 on a 2.4GHz P4, WinXP Home

Additional remarks on Colomn values:

1) For completeness I should have mentioned that my laptop is updated with SP2 and has all automatic updates from MS.

2) For verification I will attempt to include a screenshot of the colomns.
=
Sorry, jpeg paste not possible ! just ask me if you need this screenshot
=
3) As the model has entered Phase 2 of experiment 1 and P1 has generated stable graphs, the conclusion that the value errors where not caused by the model, is justified.

Worth mentioning is that during/directly after the Phase-transition 1>2 the value for "CPU time" reduced by more than 2 hours ( from 236:xx:xx to 234:xx:xx )
The "Progess" value was not affected and the "To Completion" value might have been affected, (but as I was not tracing that value I can't be sure).

The last change has effect on the Average s/Ts value, which I used as a PI for my model/machine. As seen below:

Time Sent (UTC) Host ID Result ID Result Name Phase Timestep CPU Time (sec) Average (sec/TS)
13 Mar 2005 22:18:06 116410 581232 2czn_200131263_1 2 10802 881578 3.2645
13 Mar 2005 11:57:39 116410 581232 2czn_200131263_1 1 259248 850742 3.2816
13 Mar 2005 01:08:34 116410 581232 2czn_200131263_1 1 248446 814938 3.2801

With some other client related issues, such as:

1) On opening and running the Graphics screen the cpu-share for the model falls from 99% to 50% in best cases and comes almost to a complete standstill in worst cases.

2) For no appearant reason I had to recycle my machine one time as the client was hanging.

3) Shutdown procedures, to avoid loss of many hours of computation, are clumsy rules to hide the fact that the software is still in an infant stage of development.

The risk of losing the computation hours in case there is a power outage, I have reduced by suspending the model as soon as my laptop reverts back to batteries, but in the case I run an other application which causes a total hang ( mouse and keyboard frozen), than I have the risk of losing up to 1000 hours or more when running the 100 year model in the future.

and this because of a buggy client, NO Thanks!

Neigher is upgrading to a newer client a save solution, as the reports are rolling in that users lose there work in the upgrade and other server-connection issues with the 4.25 client.


I might be negative about this client, but it is a public version. not a beta !

For a year I have been running Sun Solaris 10 beta software and encountered less problems with that than with this public client, so improvements to get a reliable client is a must.

---