The topic is driven by the very disk intense Clean Energy tasks, but applies to every BOINC project running off a Windows machine (Linux and probably *Nix* based OS X not nearly as much affected).

Long story short, watching the amount of time go by that is wasted on Disk IO decided yesterday morning to run a "defragmentation" with the free "AusLogic". There was just 2% fragmentation according the pre-analysis and some 8500 files broken up.

Started the defrag and before going on an out and about watched the progress for a while and was stumped by the endless series of slots/1/ slots/2/ slots/3/ slots/4/ files being shuffled around. This was a fast defrag run, not a defrag/optimize org run. Continued to let BOINC crunch... like to live dangerously... and let it rip whilst outdoors enjoying the sun of the continuing summer (21c).

On comeback things were running still nicely, defrag finished and let BOINCTasks log the results, mixed with Clean Water and looked in again this evening. The results, some 6 logged below at and efficiency of 97%+, which is very nice compared to the often observed and co-reported by other crunchers as more being in the range of 92-93%.

6.35 cep2 E200529_751_A.26.C20H10N2S3Si.1.4.set1d06_1 12:21:07 (12:00:00) 14-11-2010 17:01:04 14-11-2010 17:07:05 Reported: OK (u) 21 min.
6.35 cep2 E200529_766_A.26.C18H10N4S4.41.0.set1d06_0 08:40:58 (08:23:58) 14-11-2010 12:34:37 14-11-2010 12:41:45 Reported: OK (u) 17 min.
6.35 cep2 E200529_746_A.26.C19H10N2OS4.3.1.set1d06_0 11:41:37 (11:20:10) 14-11-2010 4:39:14 14-11-2010 4:46:16 Reported: OK (u) 21 min.
6.35 cep2 E200529_682_A.26.C19H10N2OS4.75.1.set1d06_1 09:57:25 (09:36:01) 14-11-2010 3:52:55 14-11-2010 4:00:00 Reported: OK (u) 21 min.
6.35 cep2 E200529_750_A.26.C19H10N2OS4.70.0.set1d06_1 08:02:39 (07:35:06) 13-11-2010 17:36:28 13-11-2010 17:43:32 Reported: OK (u) 27 min.
6.35 cep2 E200529_994_A.24.C19H12S3SeSi.11.3.set1d06_1 09:51:38 (09:29:10) 13-11-2010 16:36:55 13-11-2010 16:43:59 Reported: OK (u) 22 min.
----------------- Those before the defrag below, Those after defrag above.
6.35 cep2 E200526_445_A.25.C22H15NOS.81.0.set1d06_0 09:41:48 (08:55:52) 13-11-2010 2:38:08 13-11-2010 2:44:41 Reported: OK (u) 46 min.
6.35 cep2 E200526_440_A.25.C22H15NOS.147.3.set1d06_0 09:53:17 (08:40:52) 12-11-2010 21:52:16 12-11-2010 21:59:22 Reported: OK (u) 73 min.
6.35 cep2 E200515_673_A.25.C20H14N2OSSi.47.1.set1d06_0 09:33:59 (08:36:48) 12-11-2010 15:12:28 12-11-2010 15:32:41 Reported: OK (u) 57 min.

With the above in progress looked into the CEP2 slots and was taken aback by the > 6600 files per slot and upon doing a second defrag, found again some 7500 files had split all over and 99% all being in the CEP2 slots. Started a 3rd defrag and this time set the option to put the system files at the start, this to maximize contiguous free space. More CEP2 is crunching in meantime and monitored.

For now with this, have progressed as follows:

1. Created a logical partition of 10GB exclusively for BOINC data-dir and will move it there when no more CEP2 jobs are running (yes those multi hour checkpoint intervals make you think twice before stopping BOINC at leisure)
2. Created another logical partition to put the swap file on.

Hoped gain: Minimize the continuous fragmentation, Windows just dumping parts of files in any free block it comes across. As jobs end, BOINC will delete the slot content and thus clean out the used contiguous space and reuse for optimal performance.

I'm not the first to bring it up, but now it proofs to be really worthwhile for Windows crunchers working on big tasks. Expect some gain too in general BOINC crunching when doing other sciences too.

--//--

PS, this is not chat IMMHO :-)

edit: More results from the crofts for CEP2 in above setup. Going to let it run for a while in unmanaged mix, with the "Yes" override option, so it could end up having more than 2 running concurrently.

6.35 cep2 E200542_104_A.24.C21H14S2Se.58.3.set1d06_1 09:00:50 (08:43:54) 15-11-2010 11:19:51 15-11-2010 11:26:55 Reported: OK (u) 17 min.
6.35 cep2 E200529_784_A.26.C19H10N2OS4.99.1.set1d06_0 08:46:50 (08:29:38) 15-11-2010 4:44:15 15-11-2010 4:51:17 Reported: OK (u) 17 min.
6.35 cep2 E200529_783_A.26.C19H10N2OS4.185.1.set1d06_1 10:08:09 (09:50:08) 15-11-2010 1:06:48 15-11-2010 1:13:50 Reported: OK (u) 18 min.