Date: Wed, 17 Jun 2009 12:53:37 +0200 From: Ivan Voras <ivoras@freebsd.org> To: freebsd-fs@freebsd.org Cc: freebsd-stable@freebsd.org Subject: Re: ZFS performance on 7.2-release/amd64 low compared to UFS2 + SoftUpdates Message-ID: <h1ahvg$f3t$1@ger.gmane.org> In-Reply-To: <cf9b1ee00906170034q1cee4581hb518f53e9f368368@mail.gmail.com> References: <cf9b1ee00906170034q1cee4581hb518f53e9f368368@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Dan Naumov wrote: > I am wondering if the numbers I am seeing is something expected or is > something broken somewhere. Output of bonnie -s 1024: Unless you have 512 MB of memory in the machine or you're trying to test caching, the benchmark you did is useless. In your environment, you need at least "-s 4096". Even with those issues solved, it's semi-useless since you did both tests on the same drive, on different parts of it (see "diskinfo -vt ad0" or whatever your drive is to see how different parts of the drive have different performance). To make an objective comparison you need two identical drives, and create a new empty small-ish partition (e.g. 15 GB) on the same position on both (e.g. at the start), then use this partition only for benchmarking (not for the OS, etc). > on UFS2 + SoftUpdates: > > -------Sequential Output-------- ---Sequential Input-- --Random-- > -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks--- > Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU > 1024 56431 94.5 88407 38.9 77357 53.3 64042 98.6 644511 98.6 > 23603.8 243.3 > > on ZFS: > > -------Sequential Output-------- ---Sequential Input-- --Random-- > -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks--- > Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU > 1024 22591 53.7 45602 35.1 14770 13.2 45007 83.8 94595 28.0 102.2 1.2 I did my own testing on the early import of ZFS, the results in bonnie++ were that read and rewrite speeds are significantly better on ZFS than on UFS+SU (50%+), while write speed is a bit slower (~~10%). There are of course other workloads than the sequential that need to be reviewed. For example, blogbench places ZFS again at about 50% better than UFS+SU, while randomio makes it 50% slower. Untarring the ports tree on ZFS is about 3x faster than on UFS+SU.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?h1ahvg$f3t$1>