Date: Thu, 22 Oct 1998 10:18:14 +0930 From: Greg Lehey <grog@lemis.com> To: "Alok K. Dhir" <adhir@worldbank.org>, current@FreeBSD.ORG Subject: Re: Vinum vs CCD? Message-ID: <19981022101814.D1219@freebie.lemis.com> In-Reply-To: <Pine.BSF.4.05.9810211135070.18184-100000@shadow.worldbank.org>; from Alok K. Dhir on Wed, Oct 21, 1998 at 11:37:13AM -0400 References: <Pine.BSF.4.05.9810211135070.18184-100000@shadow.worldbank.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wednesday, 21 October 1998 at 11:37:13 -0400, Alok K. Dhir wrote: > > Hi - I recently stumbled accross the existence of vinum in 3.0, and have a > couple of simple questions: > > 1. How does vinum compare to CCD, specifically with regards to > performance and stability? We haven't done much performance testing. My indications, using dd from a raw volume, are that even with the current debug code in Vinum, it will drop performance by about 2% compared to a standard partition. The throughput of both raw partition and raw volume was in the order of 10 MB/s (this is a new IBM IDE drive). I haven't tried it with ccd, but I don't expect much difference. Usingn striping, you can improve overall performance with multiple access only, so it wouldn't show up with a simple dd. When I have time, I'll start doing some serious performance comparisons. > 2. Is anyone using Vinum on an SMP -current box with any horror/success > stories? Yes, with both :-) It seems to work fine with IDE drives, but a couple of people have reported problems in conjunction with CAM and the RAID-5 version. We haven't quite identified the causes, but it seems that under some situations CAM will return an I/O error, and Vinum doesn't always handle it correctly. I'm working on the Vinum side of this problem; the CAM side may be due to incorrect termination which only shows up in typical highly concurrent access patterns which Vinum RAID-5 causes, or in one case it's been suggested that the power supply can't handle the demands of several drives being accessed concurrently. We don't have any specific evidence of a CAM problem. Greg -- See complete headers for address, home page and phone numbers finger grog@lemis.com for PGP public key To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?19981022101814.D1219>