[e2e] on local ethernet throughput?

[Cannara]

I don't have a good scanner, David, but I'll be happy to mail or fax a copy.
The curves are indeed interesting, since they clearly show the highly
nonlinear effects of high-rate collisions at very high opffered loads, but
they also show that the equivalent Token ring starts out with a much larger
delay penalty and, despite its slower growth with load, ends up crossing
Enet's fast-rising delay to the right (at higher load) of where it had crossed
for fewer nodes.  Thus, adding nodes to Token worsens its relation to an
Ethernet serving the same nodes.

On Token Ring, you're missing the 24bit delay in every ring node that's
necessary to note a "Free" token and change it to "Busy".  In addition,
especially for the newer hub-wired rings, cable lengths were doubled, since
xmit and recv must be carried out to every node -- a doubled signal path.  So,
the very nature of Token passing is delay bounded for both mediation and
repeating.  This is the big hit that starts off well above 0 delay for any
ring and increases with the number of nodes.  In FDDI, the situation is
worsened by the need to allow long Token-Rotation times for MANs.

I agree, Man has two important capacities:  ignorance and respect.  Too much
of one and not enough of the other is often our downfall.

Alex


"David P. Reed" wrote:
> 
> At 09:33 PM 10/21/2001 -0700, Cannara wrote:
> >What Vernon says is quite right and I'll only add that Collision sensing and
> >recovery happens in times on the order of 200 uS to a few mS, on even
> >extremely highly contended Ethernet segments (many stations ready with pkts to
> >send all the time).  This is far faster than Token passing (many mS), in any
> >of its forms, as a very pertinent graph from the original IEEE 802
> >standarization simulations shows (it can be faxed to anyone who wishes a
> >copy).  This graph is particularly telling in that CSMA/CD become better in
> >relation to Token as the number of contending nodes increases -- yes, better.
> 
> I'd love to see the chart - but can you make a scanned image on a server,
> and send the URL (why fax it?).
> 
> Token rings introduced essentially one bit/node of propagation delay
> (actually slightly less), plus 8 bits of token overhead at the end of each
> packet.  At 10 Mbit/sec, that is hardly "many mS", if you have only
> hundreds of nodes on the ring (as N grows, you can make it as large as you
> like).  You are absolutely right that the corresponding "arbitration time"
> for CSMA/CD Ethernet is flat with regards to the number of nodes.  It
> actually grows with the diameter of the collision domain due to "speed of
> light", but the standard fixes it to a constant.
> 
> I suspect that is what the graph you are suggesting will show, which is why
> I'd like to see it.
> 
> It's sad to look back at these old marketing  battles and see how the
> distortions of fact arose and stuck with us.
> 
> Religion seems to work because it places much less burden on detailed
> thinking and analysis than does science.  Technical "religions" seem to
> work the same way.
> 
> - David