Date: Wed, 12 Dec 2001 09:00:34 +1030 From: Greg Lehey <grog@FreeBSD.org> To: Wilko Bulte <wkb@freebie.xs4all.nl> Cc: Matthew Dillon <dillon@apollo.backplane.com>, Mike Smith <msmith@FreeBSD.org>, Terry Lambert <tlambert2@mindspring.com>, Joerg Wunsch <joerg_wunsch@uriah.heep.sax.de>, freebsd-current@FreeBSD.org Subject: Re: RAID performance (was: cvs commit: src/sys/kern subr_diskmbr.c) Message-ID: <20011212090034.C67986@monorchid.lemis.com> In-Reply-To: <20011211153437.A69755@freebie.xs4all.nl> References: <200112101754.fBAHsRV01202@mass.dis.org> <200112101813.fBAIDKo47460@apollo.backplane.com> <20011210192251.A65380@freebie.xs4all.nl> <200112101830.fBAIU4w47648@apollo.backplane.com> <20011211110633.M63585@monorchid.lemis.com> <20011211153437.A69755@freebie.xs4all.nl>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tuesday, 11 December 2001 at 15:34:37 +0100, Wilko Bulte wrote:
> On Tue, Dec 11, 2001 at 11:06:33AM +1030, Greg Lehey wrote:
>> On Monday, 10 December 2001 at 10:30:04 -0800, Matthew Dillon wrote:
>>>
>>>>> performance without it - for reading OR writing. It doesn't matter
>>>>> so much for RAID{1,10}, but it matters a whole lot for something like
>>>>> RAID-5 where the difference between a spindle-synced read or write
>>>>> and a non-spindle-synched read or write can be upwards of 35%.
>>>>
>>>> If you have RAID5 with I/O sizes that result in full-stripe operations.
>>>
>>> Well, 'more then one disk' operations anyway, for random-I/O. Caching
>>> takes care of sequential I/O reasonably well but random-I/O goes down
>>> the drain for writes if you aren't spindle synced, no matter what
>>> the stripe size,
>>
>> Can you explain this? I don't see it. In FreeBSD, just about all I/O
>> goes to buffer cache.
>>
>>> and will go down the drain for reads if you cross a stripe -
>>> something that is quite common I think.
>>
>> I think this is what Mike was referring to when talking about parity
>> calculation. In any case, going across a stripe boundary is not a
>> good idea, though of course it can't be avoided. That's one of the
>> arguments for large stripes.
>
> In a former life I was involved with a HB striping product for SysVr2
> that had a slightly modified filesystem that 'knew' when it was
> working on a striped disk. And as it know, it avoided posting I/O s
> that crossed stripes.
So what did it do with user requests which crossed stripes?
Greg
--
See complete headers for address and phone numbers
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?20011212090034.C67986>