[lug] my "no-data-movement" file-copy attempt

Doug Pintar ratnip3 at gmail.com
Tue Oct 18 01:33:51 MDT 2011


Gang,
Thank you to all who graciously replied to my previous query.  After some further testing, all I can say is that I must have been awake too long when it thought I'd observed the "problem".  Mea culpa.  Further testing has revealed that the system performs as advertised, with inter-drive copy rates approaching the maximum transfer rate of the slower drive.  It is interesting, though, that on my Compaq Linux box (Presario SR1950NX, Athlon 64 3800+ @2.4GHz, 1.5GB PC3200 DDR RAM, BIOS v3.11 [which appears to be the current rev]) I have two Western Digital SATA hard drives, primary being a WD5000AADS, secondary a WD2500JS.  These are both 7200 RPM drives; the stated "maximum internal transfer rate" for the 500 GB drive is 108 MB/sec (and this is probably drive-manufacturer decimal MB rather than MiB) and the 250 GB drive is 93.5 MB/sec.  I see a Linux transfer rate on the big disk of about 95 MiB/sec, which is believable.  On the 250, however, I can't get it to go faster than about 58 MiB/sec.  This is independent of read or write direction.  You'd think, given the specs, that the numbers would be a lot closer.  They're both set up identically in the BIOS and have the same options in /etc/fstab.  This is all partition-independent (although the secondary drive is all one big partition).  All are ext4 filesystems ('cept swap, natch), although reading from the "raw" devices should negate any effect of that.  System is effectively idle with no CPU load or other disk activity during testing.  Once again, I'm running Mint 11 with a mostly-stock (very minor configuration changes) 2.6.38-11.50 kernel.  Any ideas?  Not a major issue by any means but simply puzzling.
Thanks again, and I'll promise to quit being a pest one of these days...
Doug Pintar
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lug.boulder.co.us/pipermail/lug/attachments/20111018/1a1a3234/attachment.html>


More information about the LUG mailing list