Message boards : Number crunching : Heavy Hard Drive Fragmentation Caused By CPDN
Message board moderation
Author | Message |
---|---|
Send message Joined: 10 Aug 04 Posts: 94 Credit: 309,849 RAC: 0 |
I noticed that there were alot of files which were fragmented on my HD several days ago (just started crunching for cpdn about 4 days ago), mainly pertaining to cpdn. I defragmented all my HD's. Today I checked again after noticing a relatively small slow down of applications on one of my computers. I did an anaylsis of all my computers (4) HD's and found them fragmented to the tune of 22-25%; 95% of which were all cpdn files, averaging well over 2,000 fragmented cpdn files. This took only three days to occur. All my <a href="http://climateapps2.oucs.ox.ac.uk/cpdnboinc/hosts_user.php?userid=760">computers</a> use XP home SP1. All HD's are 120Gb to 200Gb in size except for the laptop which is 40Gb. Has anyone got a clue why the client/app is causing this and a way to stop it? Consider this a heads up. <img src="http://boinc.mundayweb.com/cpdn/stats.php?userID=35&team=off"><a href="http://mysite.wanadoo-members.co.uk/thefinalfrontear/index.html">Team Link<a> |
Send message Joined: 5 Aug 04 Posts: 39 Credit: 14,887 RAC: 0 |
Since the fragmentation will almost purely be in CPDN files (due to the fact that they are written to periodically) this will probably not affect your other programs. After a while the files will be deleted and the spaces reused to the next few WUs, so at some point your fragmentation percentage will most likely stabalize - at least it did for me with tests on my systems. If CPDN supports it you can increase the minimum time between HD acesses that the client does. If you do this, the files will be written in larger chunks - ie. less fragmentation. This is done in your settings on the website (general preferences?). A good value is something more than 3600 (every hour or more) if your computer is stable or 360 (every 10 mins) if it crashes at times. Experiments with SETI shows that their client can run almost completely without writing to the disk if the setting is set higher than the amount of seconds it takes to complete the WU. |
Send message Joined: 10 Aug 04 Posts: 94 Credit: 309,849 RAC: 0 |
Thanks <a href="http://www.sassychero.com/wavs1/harddrive.wav">Janus.</a> I'll give it a try. <img src="http://boinc.mundayweb.com/cpdn/stats.php?userID=35&team=off"><a href="http://mysite.wanadoo-members.co.uk/thefinalfrontear/index.html">Team Link<a> |
Send message Joined: 5 Aug 04 Posts: 907 Credit: 299,864 RAC: 0 |
unfortunately the climate model does not support the boinc "write to HD time" setting. this model was written by hundreds of scientists over probably 20-30 years, so it probably wasn't the best for file i/o optimization etc. |
Send message Joined: 10 Aug 04 Posts: 94 Credit: 309,849 RAC: 0 |
> unfortunately the climate model does not support the boinc "write to HD time" > setting. this model was written by hundreds of scientists over probably 20-30 > years, so it probably wasn't the best for file i/o optimization etc. > NOT GOOD. |
Send message Joined: 5 Aug 04 Posts: 36 Credit: 2,559,795 RAC: 0 |
And if you plan on running multiple models with SMP, be perpared to give up about 5 GIG of disk space.. <img src="http://boinc.mundayweb.com/cpdn/stats.php/userID:55/trans:off/.png"> |
Send message Joined: 7 Aug 04 Posts: 33 Credit: 215,841 RAC: 0 |
I've been getting heavy fragmentation because of BOINC aswell. Even within 10 minute my drive will have 10% fragmentation from it (3gb HDD). |
©2024 cpdn.org