Date: Thu, 16 Mar 2006 14:08:20 -0500 From: Chuck Swiger <cswiger@mac.com> To: Don O'Neil <don@lizardhill.com> Cc: freebsd-questions@freebsd.org Subject: Re: Raidtest/3Ware 6000 Throughput Message-ID: <4419B7A4.8050002@mac.com> In-Reply-To: <040601c64928$c3a4e140$0300020a@mickey> References: <040601c64928$c3a4e140$0300020a@mickey>
next in thread | previous in thread | raw e-mail | index | archive | help
Don O'Neil wrote: > I would have thought I would at least see the raw single drive throughput, > plus maybe a bit more. When choosing RAID levels, you are making a tradeoff between performance, reliability, and cost. Choosing RAID-5 means you value performance the least of the three: If you prefer... ...consider using: ----------------------------------------------- performance, reliability: RAID-1 mirroring performance, cost: RAID-0 striping reliability, performance: RAID-1 mirroring (+ hot spare, if possible) reliability, cost: RAID-5 (+ hot spare) cost, reliability: RAID-5 cost, performance: RAID-0 striping If you've got enough drives, using RAID-10 or RAID-50 will also improve performance compared to stock RAID-1 or RAID-5 modes. > I've benched these drives independantly at 20+ > MB/second... Is the 3ware card really slowing things down that much with the > RAID-5 overhead? Yes. It will be less noticeable with big transactions, and more noticeable with lots of tiny ones. > What "real HW RAID-5" controller would you suggest? I'd like to stick with > IDE/ATA since I have a bunch of drives already. Maybe the 3ware 9500S -4 or -8...? > Am I maybe CPU bound, or have another issue? You're probably I/O bound, not CPU bound. -- -Chuck
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4419B7A4.8050002>