Article View: comp.arch.storage
Article #34Re: SCSI (was Re: IPI-2 questions)
From: AJ Casamento
Date: Fri, 24 Jul 1992 12:00
Date: Fri, 24 Jul 1992 12:00
74 lines
3732 bytes
3732 bytes
In article <22020@venera.isi.edu> rod@isi.edu (Rodney Doyle Van Meter III)
writes:
Rod,
I'd like to discuss a couple of your points:
>> Just to pick a nit or two, and ask a question:
>> Yes, SCSI-2 is 10 Mtransfers/sec, meaning 10 MB/s @ 8 bits, 20 MB/s @
>> 16 bits. NCR claims their 53C710 is good to 12.5 Mt, but points
>> out that it violates the standard that way.
My understanding of the SCSI-2 standard is that it calls out a standard
minimum delay of 200MHz (5 Mtransfers/sec). It also allows for the implementa-
tion of Fast Differential (I believe that Fast Single-ended is not supported
and would only allow a cable length of 0.5M) which calls out a minimum delay
of 100MHz (10 Mtransfers/sec).
The size of the transfer is started at 8bit, but can go to either 16bit or
to 32bit. This would mean that the architectural limit in the current spec
for Fast Very-Wide (ie. 32bit) Differential SCSI-2 would be 40MB/s. As I read
the current spec the use of 16bit requires two SCSI cables to be used. So, it
would seem possible (and the spec states it is) to mix 8bit and 16bit devices
on the same bus using a Host Adapter that was capable of handling 16bit.
>> The first sentenc of the foreword of standard says, "...bus
>> with up to 8 devices". Because of the way every device determines
>> whether or not it has won arbitration, IDs above 7 are
>> impossible on SCSI. Theoretically it would work if all devcies
>> on the bus were 16 bits, but nobody's building the logic
>> to handle that since the spec forbids it, and they share the
>> bus with 8-0bit devices anyway.
I know that LUN (Logical Unit Number) support is currently in the CAM spec
that DEC uses for it's CAM implementations for Ultrix and OSF/1. Given a device
that was built to support LUN (which is to say that a single controller on the
storage box was running two or more devices) that enclosure would only consume
1 SCSI ID number, but could have up to 8 LUN assignments behind it. This would
mean that you could have up to 56 SCSI devices on that one SCSI bus (assuming
that the Host Adapter is consuming one of the SCSI ID numbers).
I know that the current CAM implementation we use would allow a single system
to support up to 4 SCSI buses, 8 SCSI ID numbers per bus and 8 LUNs per SCSI
ID number (at least the way that I read the spec). Unless I've missed out on
something?
>> Does anybody have a feel for how fast you can really make
>> Asynchronous transfer run?
Nope.
>> I think VMS has an extent-based file system.
I don't know.
>> --Rod
Thanx,
AJ
**********************************************************************
* AJ Casamento "The question is not whether or *
* Digital's TRI/ADD Program not the opinions are mine; but *
* 529 Bryant Ave. PAG-2 rather, which of my personalities *
* Palo Alto, CA 94301-1616 do they belong to?" *
* 415.617.3460 *
* ajc@decwrl.dec.com *
**********************************************************************
Technical Details
Message-ID:
<AJC.92Jul24120000@thendara.pa.dec.com>Path:
arc1-ca.pugleaf.net!sparky!uunet!decwrl!pa.dec.com!granite.pa.dec.com!ajcReferences:
<BrCG6x.Jpp@acsu.buffalo.edu> <ADAMS.92Jul16014945@PDV2.pdv2.fmr.maschinenbau.th-darmstadt.de> <BrH2sz.5qM@csn.org> <22020@venera.isi.edu>Show Full Headers
Complete Article Headers
Newsgroups: comp.arch.storage Path: arc1-ca.pugleaf.net!sparky!uunet!decwrl!pa.dec.com!granite.pa.dec.com!ajc From: ajc@pa.dec.com (AJ Casamento) Subject: Re: SCSI (was Re: IPI-2 questions) In-Reply-To: rod@isi.edu's message of 24 Jul 92 01:30:39 GMT Message-ID: <AJC.92Jul24120000@thendara.pa.dec.com> Sender: news@PA.dec.com (News) Organization: Digital Equipment Corporation References: <BrCG6x.Jpp@acsu.buffalo.edu> <ADAMS.92Jul16014945@PDV2.pdv2.fmr.maschinenbau.th-darmstadt.de> <BrH2sz.5qM@csn.org> <22020@venera.isi.edu> Date: 24 Jul 92 12:00:00 Lines: 74