"IMO a value of 0 might reduce the server load a little bit more compared to 3600 b/c now the server knows earlier if enough tasks of a given work unit did return valid results and can go on w/ remaining (cleaning) tasks."
That all depends. Does reporting a result also request new work? If so, the above is NOT true if the server does has resend lost work units" enabled. A shorter back off creates a huge load on the server because every time the client connects, it sends it's entire list of WUs to to the server to have it compare against the list the server has to make sure none are missing. If the client has a fast GPU and a large cache, that could be several hundred of WUs. Granted, the bandwidth could be reduced by almost 50% if BOINC used XML attributes instead of XML elements, but that's evidently only a problem for projects like mine where I pay my own network line fees. Regardless, the bandwidth charts for Collatz directly correlate to the back-off time in the project config. If I change it from 3 minutes to 1 minute, the bandwidth and server load both triple. Jon Sonntag On Sun, Jan 12, 2014 at 5:19 AM, Toralf Förster <toralf.foers...@gmx.de>wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA256 > > Hi Dave, > > thx for 4cdc8c73e. Based on that I can now test different values - eg 0 - > for MAX_REPORT_DELAY. > SO far with a value zero the behaviour of boinc here at my system is now > the one I do expect: if a task is finished, it is uploaded, a new is > downloaded and after that the task is reported as completed: > > > $ tail -n 17 /var/lib/boinc/stdoutdae.txt > 12-Jan-2014 12:07:01 [Einstein@Home] Computation for task > LATeah0047U_1360.0_389180_0.0_2 finished > 12-Jan-2014 12:07:03 [Einstein@Home] Started upload of > LATeah0047U_1360.0_389180_0.0_2_0 > 12-Jan-2014 12:07:03 [Einstein@Home] Started upload of > LATeah0047U_1360.0_389180_0.0_2_1 > 12-Jan-2014 12:07:05 [Einstein@Home] Finished upload of > LATeah0047U_1360.0_389180_0.0_2_0 > 12-Jan-2014 12:07:05 [Einstein@Home] Finished upload of > LATeah0047U_1360.0_389180_0.0_2_1 > 12-Jan-2014 12:07:05 [Asteroids@home] Sending scheduler request: To fetch > work. > 12-Jan-2014 12:07:05 [Asteroids@home] Requesting new tasks for CPU > 12-Jan-2014 12:07:07 [Asteroids@home] Scheduler request completed: got 2 > new tasks > 12-Jan-2014 12:07:09 [Asteroids@home] Started download of input_75608_9 > 12-Jan-2014 12:07:09 [Asteroids@home] Started download of input_75608_25 > 12-Jan-2014 12:07:11 [Asteroids@home] Finished download of input_75608_9 > 12-Jan-2014 12:07:11 [Asteroids@home] Finished download of input_75608_25 > 12-Jan-2014 12:07:11 [Asteroids@home] Starting task ps_131229_75608_25_0 > using period_search version 10210 (avx) in slot 0 > 12-Jan-2014 12:07:13 [Einstein@Home] Sending scheduler request: To report > completed tasks. > 12-Jan-2014 12:07:13 [Einstein@Home] Reporting 1 completed tasks > 12-Jan-2014 12:07:13 [Einstein@Home] Not requesting tasks: don't need > 12-Jan-2014 12:07:18 [Einstein@Home] Scheduler request completed > > > IMO a value of 0 might reduce the server load a little bit more compared > to 3600 b/c now the server knows earlier if enough tasks of a given work > unit did return valid results and can go on w/ remaining (cleaning) tasks. > > - -- > MfG/Sincerely > Toralf Förster > pgp finger print:1A37 6F99 4A9D 026F 13E2 4DCF C4EA CDDE 0076 E94E > -----BEGIN PGP SIGNATURE----- > Version: GnuPG v2.0.22 (GNU/Linux) > Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/ > > iF4EAREIAAYFAlLSei4ACgkQxOrN3gB26U7GUgD/THYINEMB01PxvRjLj0ofRDXu > PjAtabeRNblA3/6B7K8A/iIzghfoav6/Jwj/J9ddaogcngdo2flnWc+/0OGFzZne > =grce > -----END PGP SIGNATURE----- > _______________________________________________ > boinc_dev mailing list > boinc_dev@ssl.berkeley.edu > http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev > To unsubscribe, visit the above URL and > (near bottom of page) enter your email address. _______________________________________________ boinc_dev mailing list boinc_dev@ssl.berkeley.edu http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev To unsubscribe, visit the above URL and (near bottom of page) enter your email address.
