Date: Sat, 10 Dec 2005 18:46:22 -0500 From: Kris Kennaway <kris@obsecurity.org> To: David Kelly <dkelly@HiWAAY.net> Cc: FreeBSD_Questions FreeBSD_Questions <freebsd-questions@freebsd.org>, Kris Kennaway <kris@obsecurity.org> Subject: Re: Freebsd Theme Song Message-ID: <20051210234622.GA83235@xor.obsecurity.org> In-Reply-To: <4FA41E1E-89C6-4687-91C7-C1A343DDCBDF@HiWAAY.net> References: <20051210172500.58401.qmail@web33302.mail.mud.yahoo.com> <439B17CA.30309@thingy.apana.org.au> <20051210201601.GB79654@xor.obsecurity.org> <4FA41E1E-89C6-4687-91C7-C1A343DDCBDF@HiWAAY.net>
next in thread | previous in thread | raw e-mail | index | archive | help
--AhhlLboLdkugWU4S Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: quoted-printable On Sat, Dec 10, 2005 at 05:34:23PM -0600, David Kelly wrote: > >But anyway, FreeBSD 6.0 is hugely superior to 5.4 and 4.11 in > >filesystem performance. I have been measuring this carefully for the > >past couple of months and hope to have the paper out soon. >=20 > For instance in 5.4 the fastest I could write to my /usr/ partition =20 > on a simple default-partitioned UDMA100 drive was 16 MB/sec with a =20 > 2.8 GHz P4 while it was capable of reading at over 40 MB/sec. Saw =20 > RELENG_6 writing on that partition at over 40 MB/sec recently. =20 > Unscientific tests using "systat -v" and moving big files. >=20 > A gvinum striped volume on two SATA150 drives routinely produces 70 =20 > MB/sec reads and writes. On this amr array with 4 disks, write performance is up to 150 MB/sec on the device with multiple processes writing to the filesystem. This makes it a good testbed because there's a lot of room to observe scaling under varying loads. > Its nice that FreeBSD is now close to the hardware's performance. > One =20 > nit is that with such a large sustained access other small accesses =20 > are starved. Probably a scheduler issue, and I'm sure the scheduler =20 > is being worked on. Yes, I've noticed that too. It also occurs in 4.11 and 5.4, but is about 50% less severe on 4.11. The ULE scheduler is much better in this respect, but processes run about 5-20% slower under most loads. Kris --AhhlLboLdkugWU4S Content-Type: application/pgp-signature Content-Disposition: inline -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.2 (FreeBSD) iD8DBQFDm2jNWry0BWjoQKURAt38AJ4ukLHmCc2jn5qbxBaYmCa9elfxTQCg+WN2 PceEAuPNzbCX70tUgmYPjdw= =p92Z -----END PGP SIGNATURE----- --AhhlLboLdkugWU4S--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20051210234622.GA83235>