Server restarted


Advanced search

Message boards : News : Server restarted

Author Message
lancone
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 26 May 14
Posts: 219
Credit: 53,963
RAC: 0
    
Message 166 - Posted: 2 Jul 2014, 7:04:35 UTC

Server has restarted. New software version in preparation to deal with hanging VMs (symptom is WUs running beyond 100% estimated time).
____________

Dennis Wynes
Avatar
Send message
Joined: 25 Jun 14
Posts: 19
Credit: 1,007,409
RAC: 0
    
Message 168 - Posted: 2 Jul 2014, 8:27:48 UTC

Good work. ;-)
I'm pleased to report a nice run of successfully completed units, so all is good.

http://atlasathome.cern.ch/ATLAS/results.php?userid=375
____________

Profile Phil1966
Send message
Joined: 14 Jun 14
Posts: 39
Credit: 1,185,758
RAC: 1
    
Message 169 - Posted: 2 Jul 2014, 8:31:01 UTC
Last modified: 2 Jul 2014, 8:34:17 UTC

Hello !

Thank you for the update.

Don't you think the "estimated time" might be "wrong" because the CPU use is only about 30 % (average) ?

Seems logical to me if BM compares the WU's size to the CPU capabilities.

Estimated time : 10 mins
Running time : 30 mins
CPU use : 30 %

FYI : CPU Time is less than 50 % of Running Time. It was not the case before.

Again, thank you for your good communication and all your efforts.

Kind Regards,

Phil1966

lancone
Project administrator
Project developer
Project tester
Project scientist
Send message
Joined: 26 May 14
Posts: 219
Credit: 53,963
RAC: 0
    
Message 175 - Posted: 2 Jul 2014, 13:58:42 UTC - in response to Message 169.

Hello,

Actually it is a bit more complicated and we have to find a better way to publish the estimated time.

The CPU efficiency observed up to now on BOINC is ~40% in average.

For each WU 10 events are processed and for a given type of event the processing time per event fluctuates (there could be very large fluctuations depending on the complexity and the topology of the events). Then in addition, different type of events (very simple to very complicated), in bunch of 10, are processed.

For very simple events (fast to process), he initialisation phase might become a non negligible component. The network connection in a critical factor for the initialisation time as the software has to be uploaded to the client. The bandwidth vary a lot from client to client.

We are working on better communication.

Regards

Profile Phil1966
Send message
Joined: 14 Jun 14
Posts: 39
Credit: 1,185,758
RAC: 1
    
Message 176 - Posted: 2 Jul 2014, 17:14:17 UTC - in response to Message 175.
Last modified: 2 Jul 2014, 17:16:08 UTC

Hello Again,

Have run some WU's this afternoon.

CPU Time (=> CPU use) improved nicely. Up to 80 % for some tasks.

Some crunchers / teammates have reported tasks running up to 99,9 % but never finishing.

Some other don't understand the huge credit difference between almost same configurations.

On this pc, http://atlasathome.cern.ch/show_host_detail.php?hostid=11, the last 6 tasks ended at around 55% (as per BM) after approx 90/95 mins.

We all understand that you are still in a developpment period,
and we are happy if our comments can help you :)

Thank You

Kind Regards

Philippe

Richard
Send message
Joined: 1 Jul 14
Posts: 1
Credit: 0
RAC: 0
Message 177 - Posted: 2 Jul 2014, 18:00:36 UTC - in response to Message 166.

I switched the application to my desk top computer that has more disc space available.
I await the application
____________

Message boards : News : Server restarted