Looking at reiser4 [LWN.net]

Looking at reiser4

Posted Aug 30, 2004 5:37 UTC (Mon) by hansreiser (guest, #24323)
Parent article: Looking at reiser4

this test seems to have been performed using a tarball that was not created on reiser4. Filesystems with sorted directories are very sensitive to whether a tarball was created on them or some other filesystem, and it makes a big difference if the readdir order is the same for both packing and unpacking. If the readdir order is different, it affects every test after that. His use of 8-10 year old hardware is also different from what we designed for: reiser4 does like to use a big CPU.

I suppose I should ask if he tested on the same partition for both filesystems, or gave ext3 the outside of the platter which is faster.

He really should do what I always do when testing products, and that is allow the makers to comment before publishing results. Everyone makes mistakes like these.

I of course tried emailing lwn but got no response.

(Log in to post comments)

Looking at reiser4

Posted Aug 30, 2004 12:37 UTC (Mon) by corbet (editor, #1) [Link]

1) The exact same partition was used for all tests - I know better than to make that sort of mistake.

2) You sent mail to me personally on the weekend. Occasionally I have to actually take a bit of time with my kids, and I might just fail to respond to weekend mail on that same weekend. You will get your response.

Looking at reiser4

Posted Sep 16, 2004 3:34 UTC (Thu) by pm101 (guest, #3011) [Link]

a) I think using a tarball created on a different filesystem is perfectly fair. If you use an installer, or a program other than tar, you're not guaranteed to have the files in your perfect presorted order in the archive. I would argue having files in a more-or-less random order is a much more fair test than having them in an order optimized for the FS. Either way, looking at the present, most of the tarballs I open were created by someone else, and not by me, and so will not come from ReiserFS.

b) I'm not sure whether the use of 8-10 year old hardware is a problem, or even accurate. In 1996, the Pentium/150 was introduced. Dual P3-450 is a bit more modern, and a single P3-450 would probably be close to the mode computer used out there today in the real world (as opposed to the uberpumped computer most nerds would use). Either way, the 4GB hard drive is much, much, much more dated than the CPU: the P3-450 was introduced in 1998. 4GB drives were common in 1995. Having a drive over 50% older than the CPU probably more than compensates for any increasing gap between the speeds of the two. Either way, whether we want the disk or the CPU hosed depends on application.

c) I think it would be a better idea, in the future, to allow reviewees comment on reviews prior to publishing. Constructive suggestion. Wish it was phrased more constructively. I don't understand why you're so antagonistic with everyone -- it's really not helping your business.