From owner-freebsd-scsi Fri Jul 23 10:48: 8 1999 Delivered-To: freebsd-scsi@freebsd.org Received: from gndrsh.aac.dev.com (GndRsh.aac.dev.com [198.145.92.4]) by hub.freebsd.org (Postfix) with ESMTP id 9EF5314C91; Fri, 23 Jul 1999 10:47:58 -0700 (PDT) (envelope-from rgrimes@gndrsh.aac.dev.com) Received: (from rgrimes@localhost) by gndrsh.aac.dev.com (8.9.3/8.9.3) id KAA88500; Fri, 23 Jul 1999 10:47:27 -0700 (PDT) (envelope-from rgrimes) From: "Rodney W. Grimes" Message-Id: <199907231747.KAA88500@gndrsh.aac.dev.com> Subject: Re: Postmark results with DPT RAID-5 In-Reply-To: <14232.30911.431405.352467@trooper.velocet.ca> from David Gilbert at "Jul 23, 1999 10:14:23 am" To: dgilbert@velocet.ca (David Gilbert) Date: Fri, 23 Jul 1999 10:47:27 -0700 (PDT) Cc: yusufg@outblaze.com (Yusuf Goolamabbas), freebsd-scsi@FreeBSD.ORG, freebsd-stable@FreeBSD.ORG X-Mailer: ELM [version 2.4ME+ PL54 (25)] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: owner-freebsd-scsi@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.org > >>>>> "Yusuf" == Yusuf Goolamabbas writes: > > Yusuf> Hi, I am trying to stress test a DPT RAID controller DPT > Yusuf> SmartRaid Ultra PM3334UW (Is there any standard mechanism or a > Yusuf> methodology available for this). Any way, I am using the > Yusuf> benchmark tool "Postmark" written by Network Appliance which is > Yusuf> available via the URL > Yusuf> http://www.netapp.com/technology/level3/3022.html. > > I configured a DPT-3334UW on a K6-II/400 machine with 4 9G SCSI > drives. I found that for creating and deleting files, it couldn't be > beat (Ports would untar 4 times faster, for instance). However, I > found that a single 9G on a 2940UW would easily beat the raw > throughput. I was dissapointed with this. This disparity is usually caused by the configuration of the stripe size or chunk factor set up for the raid array. Most default raid controller parameters are optimized for tps (transactions per second) not for aggregate data transfer rates. And an ever larger problem in optimizing raid arrays for raw throughput is the need to spindle sync the drives, with correct RPL Offsets, such that the SCSI commands and data can blast out the channel at full speed without incurring rotational delay per drive/chunk of transfer. This process requires a SCSI bus analyzer, logic analyzer and a lot of patients! -- Rod Grimes - KD7CAX - (RWG25) rgrimes@gndrsh.dnsmgr.net To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-scsi" in the body of the message