Date: Wed, 5 Sep 2001 08:56:16 +0930 From: Greg Lehey <grog@FreeBSD.org> To: David Gilbert <dgilbert@velocet.ca> Cc: Chris BeHanna <behanna@zbzoom.net>, FreeBSD-Stable <stable@FreeBSD.org> Subject: Re: [stable] Re: RAID5 Message-ID: <20010905085616.L85816@wantadilla.lemis.com> In-Reply-To: <15252.51690.121158.427612@trooper.velocet.net>; from dgilbert@velocet.ca on Tue, Sep 04, 2001 at 08:32:42AM -0400 References: <030201c1349b$a0c254b0$2aa8a8c0@melim.com.br> <20010903141721.D10812-100000@topperwein.dyndns.org> <20010904105035.C10292@wantadilla.lemis.com> <15252.51690.121158.427612@trooper.velocet.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tuesday, 4 September 2001 at 8:32:42 -0400, David Gilbert wrote: >>>>>> "Greg" == Greg Lehey <grog@FreeBSD.org> writes: > > Greg> On Monday, 3 September 2001 at 14:20:56 -0400, Chris BeHanna > Greg> wrote: >>> FWIW, I set up a 4-drive RAID-5 volume with an associated >>> journaling device on Solaris 8 that performs pretty well for home >>> directories. I wouldn't expect vinum to be any different in that >>> regard, given sufficient CPU (my Solaris volume is served from a >>> two-CPU E450, just as a reference point). > > Greg> Interestingly, CPU isn't the performance issue most people > Greg> (myself once included) assume it is. Even on my original Vinum > Greg> testbed, a 468/66, CPU usage was barely measurable. That's > Greg> probably why most hardware RAID controllers use relatively slow > Greg> CPUs. > > Well... then its an interesting issue: At least on FreeBSD, my > benchmarks show that Vinum running with Adaptec (ahc) or TekRAM (sym) > controllers outperform hardware raid controllers significantly. We're > talking same drives, same cables, same computer here. > > Not only does Vinum outperform hardware RAID 5, but it also > outperforms hardware RAID 0 and RAID 1. > > Now... these are all FreeBSD tests. Maybe the FreeBSD hardware raid > controller drivers suck? I've often heard this. I've also heard the contrary. It all depends on how you do the test, and what assumptions you make. > As an example (numbers I have readily available), my test bed was 8x > 18G Atlas IV drives (SCA if it matters) in a dedicated LVD case. I > was testing the AMI MegaRAID 1500 against Vinum running with either > an Adaptec 29160 or a TekRAM (forget model number, but it was LVD-80 > not LVD-160). > > I found that write speeds were pretty much even all around... with a > slight edge to Vinum --- 4M/s writes on the AMI, 4.5M/s writes on > Vinum. Reading, however, is where vinum dusted the competition. On > either SCSI controller, Vinum could achieve 51M/s reads compared to > 18M/s reads on the AMI. Hmm. It would be nice to know what you were testing with. > Interestingly, the figure for vinum changes with CPU speed --- it > was only 38M/s reads with a PIII/450 (100Mhz RAM) ... the 51M/s > measurement was an Athlon 750 (133Mhz RAM). So it looks like Vinum is maxing out on CPU. This sounds like bonnie, which is really not an appropriate test, though it would be interesting to know why it performed so differently with the different systems; I'd expect the results to be pretty much the same under those circumstances. Greg -- See complete headers for address and phone numbers To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20010905085616.L85816>