From mboxrd@z Thu Jan 1 00:00:00 1970 From: Hans Reiser Subject: Re: reiserfs v3 vs v4 bonnie++ benchmarks Date: Tue, 09 Mar 2004 20:24:10 +0300 Message-ID: <404DFDBA.8040006@namesys.com> References: <404DFACE.4090008@namesys.com> Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Return-path: list-help: list-unsubscribe: list-post: Errors-To: flx@namesys.com In-Reply-To: List-Id: Content-Type: text/plain; charset="us-ascii"; format="flowed" To: Mihai RUSU Cc: reiserfs-list@namesys.com Mihai RUSU wrote: >-----BEGIN PGP SIGNED MESSAGE----- >Hash: SHA1 > >On Tue, 9 Mar 2004, Hans Reiser wrote: > > > >>>V3: >>>Version 1.03 ------Sequential Output------ --Sequential Input- --Random- >>> -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- >>>Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP >>>dizzywork 1G 13459 71 62922 20 26061 6 15687 83 54593 8 191.0 0 >>> ------Sequential Create------ --------Random Create-------- >>> -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete-- >>>files:max:min /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP >>>256:64000:16000/256 1287 27 165 2 11382 77 1230 25 63 1 253 2 >>> >>>V4: >>>Version 1.03 ------Sequential Output------ --Sequential Input- --Random- >>> -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- >>>Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP >>>dizzywork 1G 16926 87 52469 15 26474 11 16759 85 55349 9 166.6 1 >>> ------Sequential Create------ --------Random Create-------- >>> -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete-- >>>files:max:min /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP >>>256:64000:16000/256 863 40 1243 16 4219 47 873 46 81 1 3980 48 >>> >>> >>> >>> >>>Interesting issues: >>>- - see that the "sequential writes" test using write with a 8192 bytes >>> >>> >>> >>> >>which of the above is this, the sequential output test? >> >> > >This one: >62922 for v3 and 52469 for v4 > > what is the amount of ram vs. the amount written? If you write 10x the ram, does the performance loss go away? > > >>>buffer is lower with v4 than with v3 (and it was like that in all tests) >>>- - both reiserfs v3 and v4 have almost same number of "little" files >>>created in both sequential or random order but... reiser4 is much slower >>>than v3 here >>> >>> > > > >>are little files = 64k? >> >> > >In my oppinion yes (I consider not a small file after 256 kbytes or so). >I dont think that "definitions" matter here, but numbers do :) Im trying >to test a setup as close as it can be to my real setup (huge FS partitions >with millions of files, web files with 32kbytes average size). > > > >>I need to take the time to read through bonnie++ so that I can really >>understand it, sigh, not today... >> >> > >That would be good couse Im not sure I completly understand it myself heh. > > > >>It does seem clear though that V4 consumes too much CPU for creates and >>writes, and that we are CPU bound for this task. After we release it >>Zam will tweak it more, and I will review the write code again. He is >>tweaking the linux profiler to collect better statistics on parent callers. >> >>Thanks for gathering this data. >> >> > >You're welcomed. > >- -- >Mihai RUSU Email: dizzy@roedu.net >GPG : http://dizzy.roedu.net/dizzy-gpg.txt WWW: http://dizzy.roedu.net > "Linux is obsolete" -- AST >-----BEGIN PGP SIGNATURE----- >Version: GnuPG v1.2.4 (GNU/Linux) > >iD8DBQFATfx9PZzOzrZY/1QRAlCqAJ9dVdk9tDbrRTmBn+oEP0NHL1RaOQCgy+ag >MHOxu6aUewDIpg1X4Jq3Y6k= >=I8Ya >-----END PGP SIGNATURE----- > > > > -- Hans