From owner-freebsd-hardware Tue Jul 13 22:49:57 1999 Delivered-To: freebsd-hardware@freebsd.org Received: from feral.com (feral.com [192.67.166.1]) by hub.freebsd.org (Postfix) with ESMTP id 430F614DFC for ; Tue, 13 Jul 1999 22:49:45 -0700 (PDT) (envelope-from mjacob@feral.com) Received: from semuta.feral.com (semuta [192.67.166.70]) by feral.com (8.8.7/8.8.7) with ESMTP id WAA08120; Tue, 13 Jul 1999 22:48:32 -0700 Date: Tue, 13 Jul 1999 22:48:27 -0700 (PDT) From: Matthew Jacob Reply-To: mjacob@feral.com To: Tim Tsai Cc: hardware@FreeBSD.ORG Subject: Re: Fibre Channel Controller [ LONG RESPONSE ] In-Reply-To: <19990714003006.A10871@futuresouth.com> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-hardware@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.org > > So, to get back to your question: I'd have to say that FC is cool, and > > will really be awesome over time, but it's just *barely* at the level you > > could put it together in your own shop with. Caveat Emptor. Man, it's > > painful to admit this, but it's true. > > That is my experience while looking for FC gear. I'd love to replace > the storage systems on our dozen or so servers with two FC storage servers > and a FC switch. I couldn't find any reasonably priced solutions. > > BTW, can your driver handle that scenario with a switch? i.e. multiple > hosts accessing one or two FC arrays (with independent RAID controller). This can be done without a switch. I've had the following private loop setup, e.g., work reasonably well for me: Vixel Hub FreeBSD i386 -> | Solaris 2.7 intel -> | FreeBSD Alpha -> | Solaris 2.7 sparc -> |-(initiator/target dev I cannot name) |-4 Disk JBOD Linux i386 -> |-Sun Photon (A5000) (up to 14 disks) |-Sun Photon (A5000) (up to 14 disks) As long as the loop isn't disturbed, it's quite stable. Reasonable 'white board' disk ownership management works. If the loop is disturbed, the 'loop id shifting' thing I referred to can occur, and like I said, that code isn't well tested. I've tried a SOCAL (Sun's SBus FC card) in this configuration with mixed results (sometimes the loop just crashes and nobody can see anybody). I've tried a JNI card in this configuration, also with some mixed results. I have also had, ha ha, this configuration: Vixel Hub Solaris 2.7 sparc-> | Solaris 2.7 sparc-> | Solaris 2.7 sparc-> | Solaris 2.7 sparc-> |- 8 Disk JBOD *not* work well (I mean, it does work well, but the drive f/w was such that too many multiple initiator accesses cause the f/w on the drive to crash requiring a power cycle to correct. Really. Very. Very. Bad. I now have a friend of mine not really speaking to me because of this. Sigh). I've not tried the Symbios FC Raid controllers for quite a while, so I can't say how well they work. I'm always *very* eager to hear what people find out- I'm mostly self-funded in a lot of this, so my h/w test matrix isn't nearly as large as I would like. The switch configuration *can* work (it's early days for the code I put in), but all that really does is join together multiple loops such that if I query the Fabric name server and get list of disks back, I can 'map' them to loop ID's such that the Qlogic knows they're not on the local loop. This picture looks like, e.g.: Brocade Switch | (initiator/target dev I cannnot name)- | Sun Photon (A5000) - | | <- FreeBSD i386 | <- FreeBSD alpha | <- Linux Alpha Vixel Hub-> | | 4 Disk JBOD -> | Sun Photon -> | -matt To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hardware" in the body of the message