Re: ULTRA ATA/100 announced

From: Jesse Pollard (pollard@tomcat.admin.navo.hpc.mil)
Date: Wed Jun 07 2000 - 07:13:52 EST


willy@thepuffingroup.com:
> On Tue, Jun 06, 2000 at 03:32:18PM +0200, scoetzee@voltex.co.za wrote:
> > this is fine, as you can always add more controllers, but theoretically (never
> > tried myself)
> > you can attach 32 SCSI 3 devices to a scsi3 bus.
>
> but i've never understood _why_. if you read the docs from, say, SGI,
> they recommend not putting more than 4 drives on a chain as they will have
> completely saturated the bandwidth. if your devices aren't high-enough
> bandwidth to saturate the bus, then why not put them on a cheaper bus
> with more controllers (like, er, IDE).

That recommendation is for 100% utilization for 100% of the time.
That is mostly for the swap device if all of it is on one controller.

8 - 15 disks is more normal, scratch/tmp, user home directories. root/usr/
var may be split up (root/usr usually is one file system, var on another).

The busier the file system the more it should be spread out, but that doesn't
mean that a second file system ( low use load) can't be put on the same
controllers.

> ok, you might argue that you only have N slots in your machine --
> there are some pretty nice quad-tulip cards available, i don't see why
> people shouldn't make quad-IDE controllers (2 busses per controller =>
> 16 devices.)

16 devices / 2 = 8 busses = 8 controllers, 8 IO addresses, I'll grant
that maybe only one IRQ.

4 controllers = 8 drives. not 16.

-------------------------------------------------------------------------
Jesse I Pollard, II
Email: pollard@navo.hpc.mil

Any opinions expressed are solely my own.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/



This archive was generated by hypermail 2b29 : Wed Jun 07 2000 - 21:00:28 EST