I begin to wonder about BOINC and climate prediction.

While my BOINC client is configured to run at most 4 applications, the record so far is 6 applications. And that is a bad thing because as the number goes up, the speed of any one goes down with the dirtying of the L3 caches on my 3.06GHz caches. So the BOINC client has a major bug.

The boincmgr process is also mixed up beyond belief: suspending a process does not suspend it. When it says a process is paused, it is not necessarily paused as it continues to accumulate time. Usually the BOINC client, as reported by the boincmgr process, thinks only four applications are running, but the excess processes are running and the boincmgr process reports the increasing times for the applications marked paused. This is one thing that the climateprediction application gets right: when a climateprediction application is marked paused, it is actually in stopped state and not accumulating time. (BTW, my client is configured to leave paused processes in memory: I have 4086 GBytes RAM, and 8192 GBytes swap space, so I am surely not running out.)

But the main reason for this post is that climateprediction applications are running up quite a bit of time and no trickles are being recorded. I did a reset project a while ago to see if that would reduce the number of client errors. And it did -- sort of. Since then, there have been no trickles at all.

Yet I have been contacting the server:

CPU type GenuineIntel Intel(R) Xeon(TM) CPU 3.06GHz
Number of CPUs 4
Operating System Linux 2.4.21-32.0.1.ELsmp
Memory 4004.7 MB
Cache 512 KB
Swap space 8001.1 MB
Total disk space 7.34 GB
Free Disk Space 2.23 GB
Measured floating point speed 517.75 million ops/sec
Measured integer speed 707.02 million ops/sec
Average upload rate 2.52 KB/sec
Average download rate 115.64 KB/sec
Results 77
Number of times client has contacted server 1341
Last time contacted server 20 Jun 2005 5:43:28 UTC
% of time client is on 98.208 %
% of time host is connected -100 %
% of time user is active 99.82 %

Latest Trickles For This Host
Time Sent (UTC) Host ID Result ID Result Name Phase Timestep CPU Time (sec) Average (sec/TS)
14 Jun 2005 23:35:44 hidden 866211 2tcv_300152685_1 2

Since I reset the project, the four climate prediction applications have accumulated the following times (in hours and minutes). Surely they should have trickled at least once each, should they not? Are the servers in trouble, or what? My machine has been contacting the climate-prediction regularly, most recently, as shown above, at 20 Jun 2005 5:43:28 UTC

876:08 hadsm3um_4.13_i686-pc-linux
846:25 hadsm3um_4.13_i686-pc-linux
684:30 hadsm3um_4.13_i686-pc-linux
520:31 hadsm3um_4.13_i686-pc-linux

The servers have not accumulated trickles since June 15. This does not mean they are lost, but will <i><b>likely</b></i> be accumulated when they get the server software working OK again. When that may be I have no idea.

Yeah, Boinc is buggy, that's a fact.
But boinc dev team is improving it almost everyday (read the dev_list and the bug base to see that), so don't be too hard with it.
I had made a little message on the seti forum about similar problems (<a href="http://setiathome.berkeley.edu/forum_thread.php?id=15279">here</a>).
The bugs are nonetheless minor, and generally don't prevent the computation of the projects :o)

The trickles server is down, and all CPDN crunchers are like you (and me :o)): waiting for the server to be back and running as soon as possible.


-----------------------------------------------
<a href="http://www.boincforum.info/boinc/">boincforum</a>