Over at the SETI/BOINC site, I\'ve read up on the new time scheduling feature that\'s new to the 4.05 core client. I\'m currently participating in CPDN and I\'ve just recently reattached to SETI@home. I\'ve set the CPDN resource share to 400 and SETI to 100. So, my machine should be crunching 80% of the time. I\'ve been monitoring BOINC fora day and a half now and I find that my client still alternates between SETI and CPDN every hour. I\'ve even tried a 200/100 split (CPDN/SETI) and it still alternates every hour. Even though the report deadline for the few SETI WUs I have is two weeks away, I would imagine that BOINC would crunch CPDN for at least 2 hours before switching to SETI.

Are there other factors affecting the time schedule that I might not be aware of, causing my client to altnernate every hour despite the resource share settings?

Thanks in advance to those who respond!

The resources split is in percentages. Unless you know of a better way to set something to use more than 100% in the end, why don't you try 80-20?
--------------------
Jordâ„¢
<img src="http://boinc.mundayweb.com/cpdn/stats.php?userID=2&amp;trans=off">

&gt; The resources split is in percentages. Unless you know of a better way to set
&gt; something to use more than 100% in the end, why don't you try 80-20?
&gt; --------------------
&gt; Jordâ„¢
&gt; <img src="http://boinc.mundayweb.com/cpdn/stats.php?userID=2&amp;trans=off">
&gt;
&gt;Jorden, doesn't 400/100 still break down to an 80/20 split? That's what my client says under the Projects tab...unless the resource share shouldn't be using values over 100 to begin with.

&gt; &gt; The resources split is in percentages. Unless you know of a better way to
&gt; set
&gt; &gt; something to use more than 100% in the end, why don't you try 80-20?
&gt; &gt; --------------------
&gt; &gt; Jordâ„¢
&gt; &gt; <img> src="http://boinc.mundayweb.com/cpdn/stats.php?userID=2&amp;trans=off"&gt;
&gt; &gt;
&gt; &gt;Jorden, doesn't 400/100 still break down to an 80/20 split? That's what
&gt; my client says under the Projects tab...unless the resource share shouldn't be
&gt; using values over 100 to begin with.
&gt; For the sake of experimentation, I'll just enter 80/20 and I'll let you know how it works out.
&gt;

Well no... it's really 100% = 100% ... it's that easy to follow. :D
--------------------
Jordâ„¢
<img src="http://boinc.mundayweb.com/cpdn/stats.php?userID=2&amp;trans=off">

Resource share is not in percentages.
E.g. if you have project A on 100 and project 2 on 200, it should give project A 33% and project B 66% of the processing time.

In Boinc version 3.x the RAC (recent average credit) also had an effect on this. If you crunched for project A for a long time and then attached to a new project B, project B would get a lot of time in the beginning to get the right balance between the projects. Maybe that is still the same in version 4.05 ?

&gt; Resource share is not in percentages.
&gt; E.g. if you have project A on 100 and project 2 on 200, it should give project
&gt; A 33% and project B 66% of the processing time.
&gt;
&gt; In Boinc version 3.x the RAC (recent average credit) also had an effect on
&gt; this. If you crunched for project A for a long time and then attached to a new
&gt; project B, project B would get a lot of time in the beginning to get the right
&gt; balance between the projects. Maybe that is still the same in version 4.05 ?
&gt;
If RAC still affects the resource share, then this would explain why SETI/CPDN alternate every hour. Due to all the problems with SETI, my RAC has dropped to 174.27 whereas my CPDN RAC is at 238.27 and dropping slowly. So perhaps in a few weeks when my RACs level out their effect on the resource share will factor out and then I can be on my way to crunching CPDN 19.2 hours a day!
<img src="http://boinc.mundayweb.com/cpdn/stats.php?userID=153&amp;trans=off">

I know it worked like that in Boinc 3.x, I'm not sure however if this changed in version 4.05. So it is more like a suggestion from my side. Maybe someone else knows for sure? :-)

Well just to keep you posted, despite changing the Resource Share to 80 CPDN and 20 SETI in the preferences, it has been an hour since CPDN resumed and it has now been preempted by SETI
<img src="http://boinc.mundayweb.com/cpdn/stats.php?userID=153&amp;trans=off">

I had a look on one of my PC's. The resource share Seti:ClimatePrediction on that one is roughly 66:33% (RAC = 807:147)

I noticed Seti is getting 3 hours every time and climate 1 hour. One time during the last day seti got about 20 minutes more. I expected more time for ClimatePrediction then it gets.
I've read somewhere on the Seti boards the timeslicing feature in Boinc needs some "polishing". Maybe it still doesn't work exactly as it should?

This is a part of the my log file:


climateprediction.net - 2004-08-31 19:22:46 - Restarting result 02fy_300028151_0 using hadsm3 version 4.03
SETI@home - 2004-08-31 19:22:46 - Preempting result 04my04aa.2277.10482.779842.254_3 (quit)
climateprediction.net - 2004-08-31 20:22:46 - Preempting result 02fy_300028151_0 (quit)
SETI@home - 2004-08-31 20:22:47 - Restarting result 04my04aa.2277.10482.779842.254_3 using setiathome version 4.03
climateprediction.net - 2004-08-31 23:22:47 - Restarting result 02fy_300028151_0 using hadsm3 version 4.03
SETI@home - 2004-08-31 23:22:47 - Preempting result 04my04aa.2277.10482.779842.254_3 (suspend)
climateprediction.net - 2004-09-01 00:22:47 - Preempting result 02fy_300028151_0 (suspend)
SETI@home - 2004-09-01 00:22:47 - Resuming result 04my04aa.2277.10482.779842.254_3 using setiathome version 4.03

Maybe this thread is interesting for you also: (?)

http://climateapps2.oucs.ox.ac.uk/cpdnboinc/forum_thread.php?id=441

More information about the resource depth they talk about is here (for Boinc 3.x):

http://homepage.mac.com/pauldbuck/site-common/glossary/letter-r.html#resource-debt


:-)

&gt; I noticed Seti is getting 3 hours every time and climate 1 hour. One time
&gt; during the last day seti got about 20 minutes more. I expected more time for
&gt; ClimatePrediction then it gets.

If you have a RAC of 800 for CPDN, your resource debt is basically non-existant for S@H since it only switched to 4.xx a day ago. Give it a week or two before you can expect any sort of rational resource sharing.

<a href="http://www.boinc.dk/index.php?page=user_statistics&amp;project=cpdn&amp;userid=355"><img border="0" height="80" src="http://355.cpdn.sig.boinc.dk?188"></a>

I have griped about this aspect of Boinc before, in other project forums.

I can see no logical reason why resource sharing should be based on RAC. This is surely contrary to what any normal user would expect. I certainly did. I presumed that when I set a share ratio, then that is what Boinc would actually do.

Now that hourly time-slicing is used in version 4, this would seem to me to make RAC even more irrelevant - even bizarre! What makes it worse, is that this is not apparently documented anywhere except in the forum threads. For an application that has resource sharing as one of it's main features, I think this should be made very clear to new users up front, otherwise it is very confusing.

&gt; What makes it worse, is that this is not apparently documented anywhere
&gt; except in the forum threads.

At least to some degree it is documented here:
<a href="http://boinc.berkeley.edu/client_sched.php">http://boinc.berkeley.edu/client_sched.php</a>

At some point it wasn't depending on RAC but the local cpu recent average debt, dunno how it works now, though.

&gt; At least to some degree it is documented here:
&gt; <a> href="http://boinc.berkeley.edu/client_sched.php"&gt;http://boinc.berkeley.edu/client_sched.php</a>
&gt;
&gt; At some point it wasn't depending on RAC but the local cpu recent average
&gt; debt, dunno how it works now, though.
&gt;
@Janus, Thanks for the link.

The info there is rather technical, and I am not sure how much of it an ordinary user would understand. I think you are correct in that we should be talking about resource "debt" rather than RAC. It would be helpful to know how one can assess this one's self, so one might understand why Boinc is allocating resource to one project rather than another - which may be contrary to "common sense".

I can see the benefit that when work is in the queue, Boinc should take into account the report deadline for any WU to ensure that it completes before expiry. However I still cannot understand the logic of the need for the current mechanism, apparent based on HISTORICAL resource allocation. Unless I am missing something crucial here, then I would still use the adjective "bizarre". Some input from a Boinc developer on this subject might be illuminating, as I would really like to understand the logic of it - even if I don't agree with it.

I am lucky to have two PCs, so I do not share CPDN with any other project, but I have upgraded to Boinc 4.05 on the other PC, in order to out the new time-slicing with P@H (when it returns from upgrading) and S@H.

Pete