[8106] in linux-scsi channel archive
Re: Recomendation HW RAID controller
daemon@ATHENA.MIT.EDU (Scott Marlowe)
Mon Feb 14 19:31:09 2000
Date: Mon, 14 Feb 2000 15:26:22 -0700 (MST)
From: Scott Marlowe <smarlowe@ihs.com>
To: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: linux-scsi@vger.rutgers.edu, linux-raid@vger.rutgers.edu
In-Reply-To: <E12KSou-0005cX-00@the-village.bc.nu>
Message-ID: <Pine.LNX.4.04.10002141518480.5222-100000@css120.ihs.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
On Mon, 14 Feb 2000, Alan Cox wrote:
> > Has anyone done any benchmarks that compare Linux software RAID to hardware
> > RAID? The parity generation numbers that go by in my software RAID boxes
> > are quite impressive, so I'd guess that software RAID would hold its own,
> > but don't have any real world data to confirm or deny my beliefs.
>
> That is too simplistic. The software raid numbers are using more CPU power.
> If your box is not CPU bound (or you can add an extra CPU easily) then it
> may well be software is faster. It isnt just a straight compare though, you
> need to ponder CPU usage
True, very true, but if the CPU usage is <10% of the total CPU horsepower, I
would gladly give that up for a 2x or better performance increase over the
hardware cards I've seen for some applications.
If the numbers that go by during boot about parity generation are realistic,
then the additional load on my CPUs to keep up with a MegaRAID would be
small enough that the extra speed should make the machine, on average,
faster, since much of what I do is I/O bound, not CPU bound.
Guess it's time to borrow a stack of SCSI drives ;^)
-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.rutgers.edu
