The way I read it, the application generates output which the application
sends to the server, but the output data is sent AFTER the task is marked
as complete, so a work request is generated before the task has completely
sent the output to the server. Please note that the APPLICATION is sending
the output to the server, not the client.
I agree that if there is no output data to send, the work request and the
report should be in the same connection. However, there is a problem with
the application behavior as well.
jm7
Nicolás Alvarez
<nicolas.alvarez@
gmail.com> To
Sent by: [email protected]
boinc_dev-bounces cc
@ssl.berkeley.edu
Subject
Re: [boinc_dev] Multiple sched
09/14/2009 01:34 requests with no output files
PM
El Lunes 14 Sep 2009 14:23:11 [email protected] escribió:
> I thought that there was. It was just that the application uploaded it
> instead of the BOINC client, or sent the final data to the server instead
> of waiting for the client to do it. If this is true, it is a race to see
> whether the file is uploaded or the work request is sent first.
If the app was uploading a file, then it wouldn't have terminated yet, so
the
client wouldn't be doing a work request.
> For non-CPU intensive tasks, the report and the request should be in the
> same packet.
I don't think it's relevant whether it's non-CPU-intensive or not.
If the task has output files, currently the client often requests work
while
uploading the files, and shortly afterwards makes another request to report
the task. It's debatable whether it should wait a few seconds for the files
to upload before doing a request (asking for work and reporting in one go),
because the CPU may be idle meanwhile, etc.
But if the task has no output files, there is no argument or gray area.
It should always do both things at once (regardless of cpu-intensive-ness).
The task was already ready to report, but the client triggered a work
request
immediately before noticing there was something to report (and by
'immediately' I mean both things happened during some code with no time
delays
other than CPU speed).
--
Nicolas
_______________________________________________
boinc_dev mailing list
[email protected]
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
[email protected]
http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.
