Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 25 Mar 1995 18:23:12 -0800 (PST)
From:      "Rodney W. Grimes" <rgrimes@gndrsh.aac.dev.com>
To:        teren@lyria.stanford.edu (Terry Lee)
Cc:        freebsd-hackers@FreeBSD.org
Subject:   Re: Why IDE is bad
Message-ID:  <199503260223.SAA08801@gndrsh.aac.dev.com>
In-Reply-To: <Pine.BSF.3.91.950325180639.8842A-100000@lyria.stanford.edu> from "Terry Lee" at Mar 25, 95 06:13:14 pm

next in thread | previous in thread | raw e-mail | index | archive | help
> 
> On a SCSI bus, I can have multiple drives and if I balance the load on 
> the drives then I am able to increase the overall throughput of my disk 
> subsystem up to the theoretical max of 10 MB/s right?

Pretty much true, except you'll never get to 10MB/sec due to SCSI bus
over head (device arbitration time).

> Is this true for IDE or E-IDE?  E-IDE can achieve bus throughput > 10 
> MB/s but there are few drives that can sustain such transfers.  But if I 
> have two drives on the same IDE adapter and I balance the load across the 
> two drives, will I get the same performance benefit as with multiple SCSI 
> drives?

Yes, and the bus arbitration time is much shorter for E-IDE.

> What if I have two drives on two different IDE adapters?

Or two drives on two scsi controllers.  You simply move the arbitration
up one level doing this.  ISA bus would be about the same, PCI would
be much faster.

This last one is the move towards what mainframe systems do.  A IBM
I/O channel is only 3MB/sec, but you have many channels in a
typical system.  One channel to each DASD controller, often you only
put 1 or 2 drives on a controller (so you don't saturate the channel).

You often see 8 or 16 channels just for DASD use.


-- 
Rod Grimes                                      rgrimes@gndrsh.aac.dev.com
Accurate Automation Company                   Custom computers for FreeBSD



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199503260223.SAA08801>