[3266] in linux-scsi channel archive

home help back first fref pref prev next nref lref last post

Re: Mylex 960?

daemon@ATHENA.MIT.EDU (Eric Youngdale)
Sun Feb 8 20:57:35 1998

Date: 	Sun, 8 Feb 1998 21:18:50 -0500 (EST)
From: Eric Youngdale <eric@andante.jic.com>
To: "Leonard N. Zubkoff" <lnz@dandelion.com>
cc: jbrown@tanstaafl.busprod.com, linux-scsi@vger.rutgers.edu
In-Reply-To: <199802051949.LAA16436@dandelion.com>


On Thu, 5 Feb 1998, Leonard N. Zubkoff wrote:

>   Date: 	Tue, 3 Feb 1998 16:41:08 -0600 (CST)
>   From: "Justin Brown W." <jbrown@tanstaafl.busprod.com>
> 
>   We have a 960 dual channel ultra wide, it's QUICK.  We're using it in a NT
>   (ugh) machine right now with a single 18 gig raid 0.  Although not a
>   perfect method of benchmarking, it took less than 30 seconds to completly
>   format the drive (ntfs) with sector testing.  We actually did it a bunch
>   of times out of sheer disbelief.  I can't wait for a working 960 linux
>   driver.
> 
> While I do believe the DAC960 is quick, that sounds like a bit too quick.  If
> NT actually read all 18GB, that would imply an I/O rate of about 600MB/second.
> There's simply no way that two Wide UltraSCSI channels can do that, let alone
> the disks.  Not to mention that there isn't enough PCI or main memory
> bandwidth to sustain that.

	Yes, I agree.   Putting down a new filesystem tends to really only
involve putting down whatever metadata is required for tracking inodes
(i.e. an inode table under Unix, and probably some equivalent concept for
NTFS), putting down a root directory, plus a few other housekeeping tasks.
There is no reason for it to read/write the entire disk.

-Eric



home help back first fref pref prev next nref lref last post