Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!clyde.concordia.ca!rutgers!tut.cis.ohio-state.edu!ucbvax!hoptoad!tim
From: tim@hoptoad.uucp (Tim Maroney)
Newsgroups: comp.protocols.appletalk
Subject: Re: Enhanced LocalTalk (was: Mac Booster Modules)
Message-ID: <9342@hoptoad.uucp>
Date: 19 Dec 89 02:13:52 GMT
References: <6662@imag.imag.fr> <24772@cup.portal.com> <9241@hoptoad.uucp> <526@wcc.oz>
Reply-To: tim@hoptoad.UUCP (Tim Maroney)
Organization: Eclectic Software, San Francisco
Lines: 92

bordier@imag.imag.fr (Jerome Bordier) writes:
>> >>In November BYTE (p.219), Tom Thompson states that "the faster
>> >>DaynaTALK and FlashBox modules can coexist (with Macs not so
>> >> equipped)
  
>In article <9241@hoptoad.uucp>, tim@hoptoad.uucp (Tim Maroney) writes:
>> Yes, it is true, at least of FlashTalk.  FlashTalk Macs can coexist
>> with LocalTalk Macs.

In article <24772@cup.portal.com> MacUserLabs@cup.portal.com (Stephan
- Somogyi) writes:
>> >Non-enhanced nodes usually can't see when enhanced nodes are
>> >transmitting, and will therefore step all over enhanced traffic. This,
>> >obviously, messes things up a bit.
 
Timbo:
>> A rather vague and misleading phrase.  Yes, LocalTalk Macs will step on
>> FlashTalk packets, requiring retransmission.  However, this does not
>> cause any serious problems, other than some performance degradation.

In article <526@wcc.oz> tom@wcc.oz (Tom Evans) writes:
>Some? SOME!!!??? How about 95% degradation being "some"?

Brent Noorda did measurements on FlashTalk at TOPS that were meant for
internal distribution, rather than press releases.  They did show some
degradation, but nowhere near this magnitude.

>I've had to cope with a network that was having "some" packets stepped
>on, and it wasn't pretty. Let's put some real numbers to it.
>
>Machine "A" and "B" communicating using ATP - file copy operation at
>say 20kbytes/sec. This corresponds to 40 packets/second, all maximum
>length, and taking up 30% of the FlashTalk bandwidth at 700,400 bits/sec.
>We now talk from a slow Mac to a LaserWriter, neither of which can see
>the high speed traffic and stands a 30% chance of killing a fast packet
>per slow packet. We're running at a reasonable PAP rate of 10 packets
>per second. We are now killing Fast packets every 0.3 seconds or so.
>This incurs a 2 second timeout-retry from ATP. We're now sending 12
>fast packets in 0.3 seconds and getting stomped for 2 seconds instead
>of sending 40 per second. 12/2.3 : 40/1 is an 87% slowdown. Some?

Are these real figures or just some that seem correct?  Have you
actually done these timings?  Not meaning to seem snide; it's just that
you didn't say.  I don't think "it seems right" is any basis for
conclusions about network speed; the real world is never quite the way
we imagine it will be.

But let's grant your figures for the sake of argument.  So what do you
intend to do about it?  You'll get pretty much the same result if
anything is talking to a LaserWriter, because the LaserWriter is going
to be stepping on packets.  (I'm not sure, but I assume that a
FlashTalk-equipped Mac will avoid stepping on LocalTalk packets even
when it has to run at LocalTalk speeds.)  It doesn't matter all that
much whether the Mac has FlashTalk or not.  In any case, the obvious
solution is to improve your timeouts; make them 0.5 seconds and post
the measurements, s'il vous plait.

>Let's say the packet that gets clobbered is an ATP TREL packet (the 
>ATP XO sequence is TREQ, TRSP, TREL). Between 10% and 30% of our packets
>are TREL's (but they're shorter than the data packets - about 30 bytes).
>For CAP, each TREL dropped incurs a 30 second stall - not a nice thing
>to witness. With AppleShare, you have two listeners on the socket, so
>you can drop ONE TREL per 30 second period without penalty. Or else
>you get the difference. With this added, you're probably up to 90% to
>95% performance degradation.

Again, these timeout values are tuned badly.

>Jerome's distributor knew what he was talking about.

Not really.  You'd have pretty much the same problems with any situation
where FlashTalk-equipped Macs had to talk with LocalTalk-only devices,
like LaserWriters and many network gateways and bridges.

>Maybe the real advantage is that the physical cable can now handle THREE
>times the volume of traffic. Like three times as many Macs performing
>simultaneous data transfer before the net clogs up. Think of it as the
>difference between a one-lane and a three-lane road. Not much
>difference at 2:00am (when programmers go to work :-), but a lot of
>difference to us ordinary folk during the rush hour.

Kind of a strange thing to say right after complaining about the number
of collisions.  I hope those packets have air bags.
-- 
Tim Maroney, Mac Software Consultant, sun!hoptoad!tim, tim@toad.com

"A book is the product of a contract with the Devil that inverts the Faustian
 contract, he'd told Allie.  Dr Faustus sacrificed eternity in return for two
 dozen years of power; the writer agrees to the ruination of his life, and
 gains (but only if he's lucky) maybe not eternity, but posterity, at least.
 Either way (this was Jumpy's point) it's the Devil who wins."
	-- Salman Rushdie, THE SATANIC VERSES