Draft Minutes of T10 Plenary Meeting #20 - 3/13/9

Norm Harris x2230 nharris at eng.adaptec.com
Wed Mar 26 09:46:21 PST 1997

* From the SCSI Reflector (scsi at symbios.com), posted by:
* Norm Harris x2230 <nharris at eng.adaptec.com>
Jim Mcgrath wrote:
> * From the SCSI Reflector (scsi at symbios.com), posted by:
> * jmcgrath at qntm.com (Jim Mcgrath)
> *
>      John,
>      Thanks for the clarification - I have no problem with this, and think
>      the focus of the meeting is appropriate (I basically wanted to see if
>      I should attend or someone else).
>      One thing was discussed that could be interpreted as a protocol issue
>      and could contribute to the driver discussion is an attempt to
>      introduce a false (i.e. not counted) transistion at the beginning of
>      a burst to overcome the problem of cleanly switching the lines for the
>      first transition after a long period of no transitions.  I'd hope that
>      the working group will consider ideas like that as well as purely
>      physical layer changes, since I'd like to avoid a transceiver change
>      if some minor work in the protocol could fix the problem.
>      Jim
> ______________________________ Reply Separator _________________________________
> Subject: Re: Re[2]: Draft Minutes of T10 Plenary Meeting #20 - 3/13/9
> Author:  John Lohmeyer <John.Lohmeyer at Symbios.COM> at SMTP2
> Date:    3/24/97 10:23 AM
> * From the SCSI Reflector (scsi at symbios.com), posted by:
> * John Lohmeyer <John.Lohmeyer at Symbios.com>
> *
> At 05:50 PM 3/21/97 -0800, Jim Mcgrath wrote:
> >
> >     I am a bit confused on this upcoming meeting - since Ultra 3 could
> >     very well have protocol changes as well as speed changes, is this
> >     the scope of the meeting?  Or is it only to focus on the narrow
> >     question of Fast 80?  (It is important since different people
> >     would come depending on the scope of the meeting).
> Jim,
> The April 18th meeting in San Jose was authorized as a SPI-2 Working Group
> meeting.  Obviously, a major consideration of this meeting is whether or
> not we should recommend to T10 that SPI-2 be 'unstabilized' in the area of
> LVD drivers, receivers, and terminators.
> At the March 10th SPI-2 working group meeting, Adaptec presented data that
> they believe implies we may eventually want to be using symmetrical drivers
> with non-biased terminators and dual receivers (one biased and one not).
> SPI-2 currently specifies asymmetrical drivers with biased terminators and
> non-biased receivers.
> A key technical point that I think the working group needs to address is at
> what speed will the symmetrical drivers be needed (if at all)?
> If the case for eventually needing symmetrical drivers is persuasive, then
> a marketing recommendation is needed on the timing; should we recommend
> changes to SPI-2 or wait for a future version of SPI?  This question is
> much more difficult because many elements must be considered.  For example,
> what factors are most important to the parallel SCSI market: higher speed,
> reduced overhead, or increased connectivity?  You presented interesting
> (and controversial) data on this point at the last SCSI Working Group meeting.
> I will be preparing a draft agenda and distributing it to the reflector
> soon.  I think it is beyond the scope of the meeting to invent Fast-80 or
> 'Ultra3' (whatever Ultra3 means).  However, we will have to have some idea
> of where we are going to judge the merit of Adaptec's proposal.
> Since the meeting is only one day, I want to devote as much time as is
> necessary to the question of symmetrical vs. asymmetrical.  We owe the
> industry resolution of this question ASAP.  If any time remains, I want to
> cover other SPI-2 issues, particularly those than can benefit from the flock
> of physical experts I am expecting.
> Exploring protocol enhancements is probably not going to receive much time
> (but is extremely interesting to me).  We have agreed to put this topic on
> Tuesday afternoon of the SCSI Working Group.
> John
> --
> John Lohmeyer                 E-Mail: john.lohmeyer at symbios.com
> Symbios Logic Inc.             Voice: 719-533-7560
> 4420 ArrowsWest Dr.              Fax: 719-533-7036
> Colo Spgs, CO 80907-3444    SCSI BBS: 719-533-7950 300--28800 baud
> *
> * For SCSI Reflector information, send a message with
> * 'info scsi' (no quotes) in the message body to majordomo at symbios.com
> *
> * For SCSI Reflector information, send a message with
> * 'info scsi' (no quotes) in the message body to majordomo at symbios.com

Adaptec will go as far as saying that asymmetrical design is a bad idea
at any frequency.  If you even consider the need to switch to
symmetrical without offset for the future, you need to do it now or live
with the consequences.  Once you populate the field with asymmetry,
these devices will not work on a symmetrical system.  

There are other possible approaches to this problem, including placing
the offset switching mechanism in the terminator itself.  I believe T10
briefly considered this approach.  Adaptec has no intention of trying to
dictate the  symmetrical approach requiring switchable offset in the
receiver.  We offer this as a possible solution and have studied it to a
level of detail to believe that it is a quite feasible design.  We
submit that there are numerous system design issues that impact OEMs
that will integrate this technology.  These issues accrue from our
choice to require dual threshold voltage distribution in the system.
We also believe that there is an ASIC power premium associated with the
use of asymmetric drivers.  This only gets worse of course at higher
frequencies.  Whether most people will admit it or not, they're running
hotter than their theoretical calculations suggested.  Anything to
reduce ASIC power should be goodness.

I agree that we do want to devote the majority of time on the 18th to
the question of asymmetrical versus symmetrical.

Jim's analysis as you point out is both interesting and controversial.
Adaptec for one is not ready to sign up for a phase out strategy for
parallel SCSI.  Whether more speed/connectivity etc is needed is not the
point.  The point is that it will be availible from other interfaces and
the competitive nature of the market will swallow it up in the "keep up
with the Jones" frenzy.  At 100mb/sec it will be awful tempting for a
lot of parallel SCSI users to jump ship to serial interfaces.  So far
only one major player has been in print with their intention to ride
parallel SCSI into the sunset. Everyone else is in wait and see mode.  I
don't want to give these folks a good reason to go elsewhere if I can
keep them happy with parallel SCSI.  Hence it's important that we have a
continuous roadmap for the future.


Norman H. Harris
Staff Advisory Engineer
Principal Member, T10/T11
Adaptec Inc. 
Mission Critical Subsystems Group
619 S. Milpitas Blvd.
Milpitas, Ca. 95035
PH: (408) 957-2230
* For SCSI Reflector information, send a message with
* 'info scsi' (no quotes) in the message body to majordomo at symbios.com

More information about the T10 mailing list