Date: Tue, 26 Mar 2002 16:30:21 -0500 (EST) From: David Miller <dmiller@sparks.net> To: "Kenneth D. Merry" <ken@kdm.org> Cc: freebsd-hardware@FreeBSD.ORG Subject: Re: RAID: Adaptec vs Mylex Message-ID: <Pine.BSF.4.21.0203261617020.67145-100000@search.sparks.net> In-Reply-To: <20020326135045.A4301@panzer.kdm.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, 26 Mar 2002, Kenneth D. Merry wrote: > On Tue, Mar 26, 2002 at 15:30:56 -0500, David Miller wrote: > > Any input from real world users on the subject? I'm most interested in > > the 3210S and 352. My use would be with RAID 10 support of a busy > > database server doing zillions of writes/updates. Speed writing small > > blocks is of the essence. > > > > Reliability, robustness, speed are critical factors. It will be hooked up > > to 10 or 12 15K drives. > > > > Input? > > If you want a comparison of the two controllers: > > http://www.adaptec.com/worldwide/product/markeditorial.html?prodkey=3210_wp&type=Common&cat=%2fCommon%2fRAID+Upgrade Mylex has one on their site: (gawd frames are awful) http://www.mylex.com/products/index.html -> competitive analysis -> AcceleRAID 352 vs Adaptec 3200S (choose PDF). Naturally, it show the mylex adapter way out front. > The only catch is that the comparison was done with RAID-5, not RAID-10. Unfortunately, that makes is pretty useless for me. The two analysis look pretty far apart to start with, but when you look at the details they may complement each other pretty well. Mylex wouldn't call attention to it, of course, but area Adaptec beat it in pretty thoroughly was with big blocks, espcecially with sequential IO and write back enabled. That area is what the adaptec benchmark homes right in on, so they're not in violent disagreement. Other differences were in the disk setup - adaptec used 8 drives on four channels. Mylex used 18 drives on two channels. It may be that Mylex can handle more total commands, but if you don't have enough drives it just doesn't matter because the disks are the bottleneck. What I'm trying to sort out is whether there would be any real world performance difference to me in using either of them when updating 20 million 60 byte records in a 160 million row table. > So the read speed in the benchmarks above may be comparable to RAID-10, > but the write speed for RAID-10 should be better. > > If you want high performance with lots of small chunks of data, I would > recommend the Adaptec 5400S. It has a hardware parity engine that speeds > up RAID-5 writes significantly. (If you're only doing RAID-10, though, it > won't have any effect.) I think I really need the performance of raid 10. The database is small enough that space efficiency is not an issue. Lots of little IO's - mostly O's - are what really counts. Occasionally, handling directories with 10K entries is an issue, so I'll probably want write-back enabled. > (disclaimer: I work for Adaptec.) Full disclosure, but not necessary around these parts. Your reputation preceeds you Ken:) --- David To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hardware" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.21.0203261617020.67145-100000>