I hope I haven't lost 50 hours of work with a model because I d/l some SAH2 units?

I got the above message when SETI@home finally sent some units. BOINC stopped computing for CPDN and I get this message when I hit the CPDN update button:

climateprediction.net - 2004-08-29 23:07:46 - Deferring computation for result 00v7_300026108_0

CPDN and SAH2 are meant to share my computer. How does that work?

This is normal.

The BOINC ver 4.xx uses pre emptive scheduling. This allows BOINC to switch projects at one hour intervals. (possibly longer depending on resource share)

You are just seeing S@H Pre empt CPDN. Nothing to worry about. :)

<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>

cpdn will start back up when the seti time matches the proportion you have allocated.

for example of your "Resource Share" here at CPDN is set for 100, and your "Resource Share" over at the seti pages is also 100, they will try an do the same amount of work (CPU Time). So that may be a seti workunit completed and 1 trickle of CPDN completed.

Just be careful not to set CPDN share too low if you also have "Leave suspended tasks in memory" to "False" (the default setting), as if CPDN gets too short a time to run it will never make it to the next checkpoint (144 timesteps, which is usually 10-15 minutes).

&gt; cpdn will start back up when the seti time matches the proportion you have
&gt; allocated.
&gt;
&gt; for example of your "Resource Share" here at CPDN is set for 100, and your
&gt; "Resource Share" over at the seti pages is also 100, they will try an do the
&gt; same amount of work (CPU Time). So that may be a seti workunit completed and
&gt; 1 trickle of CPDN completed.
&gt;
&gt; Just be careful not to set CPDN share too low if you also have "Leave
&gt; suspended tasks in memory" to "False" (the default setting), as if CPDN gets
&gt; too short a time to run it will never make it to the next checkpoint (144
&gt; timesteps, which is usually 10-15 minutes).
&gt;
How long it take to get to the next check point is highly dependent on how fast the computer is. One of the computers the CPDN let in is too slow to get to the next checkpoint in an hour, and it just restarts over and over and over. The solution is to have BOINC only switch at checkpoint complete. This also avoids the problem of loosing 5% of the dedicated processing time. Assume that it takes about 112 minutes between checkpoints on average. Half of that time is 6 minutes or 5% of the CPU time lost per hour on average as the task is set back to the last checkpoint once an hour.

jm7

Better safe than sorry, so one hour for CPDN/BOINC will for sure bring me to the next timestep?

I haven't had any trickles sent since yesterday 29th August 18:38:02 UTC. Prefs are 100 for CPDN/BOINC and 200 for SAH2, which is 1 hour to the first and 2 hours for the latter. It's easy to add time for each project. For instance 200 for CPDN and 400 for SAH2 should double the periode for each project. Would that be safer?

I am a CPDN newbie so please excuse stupid questions about timesteps, trickles, and other never heard of expressions.

I 'used' to have a trickle roughly every 8 hour, when I ran CPDN/BOINC 100%. With 1/3 of my cpu power there should be a trickle for me every 24 hour, or am I completely lost her. :-)

&gt; Better safe than sorry, so one hour for CPDN/BOINC will for sure bring me to
&gt; the next timestep?
&gt;
&gt; I haven't had any trickles sent since yesterday 29th August 18:38:02 UTC.
&gt; Prefs are 100 for CPDN/BOINC and 200 for SAH2, which is 1 hour to the first
&gt; and 2 hours for the latter. It's easy to add time for each project. For
&gt; instance 200 for CPDN and 400 for SAH2 should double the periode for each
&gt; project. Would that be safer?
&gt;
&gt; I am a CPDN newbie so please excuse stupid questions about timesteps,
&gt; trickles, and other never heard of expressions.
&gt;
&gt; I 'used' to have a trickle roughly every 8 hour, when I ran CPDN/BOINC 100%.
&gt; With 1/3 of my cpu power there should be a trickle for me every 24 hour, or am
&gt; I completely lost her. :-)
&gt;
Whether you reach the next check point is entirely dependent on your processing speed. BOINC checks about once an hour no matter what you have set for your resource shares. You can set BOINC so that is suspends in place rather than exiting. This is available on the web page for your account under general settings. Suspending rather than exiting will keep you from going back to start when the process is pre-empted.

jm7

&gt; You can set BOINC so that is suspends in place
&gt; rather than exiting. This is available on the web page for your account under
&gt; general settings. Suspending rather than exiting will keep you from going
&gt; back to start when the process is pre-empted.

I did that at first but found that this solution consumed a lot of virtual memory, so programs took an eternity to start. I just have 512 mb RAM on my 3200 XP+ pc. Default is that 75% memory is dedicated to BOINC. Isn't that an awful lot?

If I loose rather much work by this swapping between projects: isn't there a way to 'swap' manually. I mean is there a way to exit ONE project without loosing the current unit/model and then restart this project 24 hours later while exiting the second project and so on?

Sorry about the double post.

&gt; Default is that 75% memory is dedicated to BOINC. Isn't that an
&gt; awful lot?
The memory setting there is virtual memory, thus swap space, not the actual memory. If you check in task manager, you'll find that both the CPDN programs take up around 55MB of normal memory plus around that same in virtual memory, while the actual work unit can take up to 600MB of virtual memory (swap space).

Yet if you don't like the setting, set it down to 25% or 50%, or add to your page file if you have the room on your harddrive. I also have 512MB of normal memory, but my page file can grow from 768MB minimum to 4095MB maximum.
--------------------
Jordâ„¢
<img src="http://boinc.mundayweb.com/cpdn/stats.php?userID=2&amp;trans=off">