User: Password:
|
|
Subscribe / Log in / New account

Reiser4 - the mammoth arrives

Reiser4 - the mammoth arrives

Posted Oct 31, 2002 18:12 UTC (Thu) by himi (guest, #340)
In reply to: Reiser4 - the mammoth arrives by cpeterso
Parent article: Reiser4 - the mammoth arrives

Reiserfs was a bit slower than ext2, but not by that much - on bonnie++ I saw numbers like ~19MB/s reads from ext2, and about 16-17MB/s reads from reiserfs. This is on a disk that hdparm -t lists as giving about 21MB/s from the platters. I can't remember about writes, but I think it was slower there too - I may have had tail packing enabled, though,which would screw with the numbers.

The real test will be how it performs with a) lots of small files, and b) big files: reiserfs has always been better at dealing with lots of small files than ext2 (which isn't saying much, and with directory indexing ext3 looks to be as fast), but it's had problems with very large files, and it's tended to have problems with fragmentation leading to performance degredation over time. If they've dealt with those problems, then I could see "100% faster" being a reasonable description, assuming of course it /is/ that much faster.

Of course, I'm not going to trust my data to Reiser4 until it's been in real use for a while . . . Maybe 2.6.5 or so ;-)

himi


(Log in to post comments)

System CPU usage

Posted Nov 1, 2002 13:02 UTC (Fri) by jzbiciak (subscriber, #5246) [Link]

One thing I've noticed in many benchmarks (well, at least the few that report this figure) is that Reiser3 seems to be very CPU heavy as compared to ext2. They may achieve about the same benchmark numbers as other filesystems when run with a disk-intensive benchmark, but (as I recall) the system CPU %age is much higher for Reiser3 than others. The effect of higher system CPU usage is to slow down any compute-intensive processes that might be running in parallel with your I/O intensive application.

I'd like to see how Reiser4 fares. It's apparently a major rewrite, so if it has fundamental changes in design, it could have completely different CPU usage patterns.

--Joe


Copyright © 2017, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds