Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 01 May 2000 19:17:01 -0700
From:      Mike Smith <msmith@freebsd.org>
To:        Parag Patel <parag@cgt.com>
Cc:        dg@root.com, Mike Smith <msmith@FreeBSD.ORG>, freebsd-stable@FreeBSD.ORG
Subject:   Re: How good is AMI MegaRAID support? 
Message-ID:  <200005020217.TAA04945@mass.cdrom.com>
In-Reply-To: Your message of "Mon, 01 May 2000 18:01:26 PDT." <99363.957229286@pinhead.parag.codegen.com> 

next in thread | previous in thread | raw e-mail | index | archive | help
> 
> [Re: BSOD and system crashes, UPS and NVRAM]
> 
> What if the RAID card's onboard software crashes?  I don't know enough
> about these cards to guess at what's more likely - FreeBSD crashing or
> the onboard card software crashing. 

It's reasonably uncommon - I've only run into a very small handful of 
cases where I've even *suspected* that I've crashed the card firmware.

> Wouldn't any crashes that would toast a filesystem under FBSD care if
> the underlying hardware was RAID or a single disk?  Wouldn't it still be
> toast either way?

The issue here is that, in order to achieve decent performance, you need 
to de-order writes.  If you assure that writes are ordered going to 
stable storage, you can maintain filesystem integrity.  If "stable 
storage" is relatively fast (eg. BB cache), you get a performance win.  
If "stable storage" is relatively slow, you get a performance loss that 
you can only trade off by de-ordering writes and thus compromising 
reliability. 

> I can see how a NV write-buffer could speed up write performance, but
> only if the disk writes are somewhat bursty.  It doesn't seem to be very
> different from disks with large onboard caches.  (Which presumably might
> still require a UPS to ride out crashes.)

The real utility of the buffer lies in being able to sort and coalesce 
write operations, as well as reducing the bus traffic.  The latter is 
actually fairly important; in a 3-disk RAID5 array you transfer 100% data 
to an embedded controller, but you would transfer 150% data if you had 
already computed the parity and were just sending it to disk.  (This 
picture gets better with more disks, but mirrored disks stay at 200%.)

-- 
\\ Give a man a fish, and you feed him for a day. \\  Mike Smith
\\ Tell him he should learn how to fish himself,  \\  msmith@freebsd.org
\\ and he'll hate you for a lifetime.             \\  msmith@cdrom.com




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?200005020217.TAA04945>