Segmentation

The only part I'm confused on is the RI/RO structure, is it:

A:    A line from Port #1 ( or whichever port ) on the switch, to Port #1
 or whichever port ) on MAU#1, and simply wrapping RI/RO cables on however
many MAU's there are on that particular ring, or:

B:   A line from Port #1 ( or whichever port ) on the switch, to RI on
MAU#1, and a line from RO on the final MAU to Port #2 ( or whichever port )
on the switch to form the ring.

The main reason for my confusion is i guess, that I thought communication
between the switch and the MAU(s) requires a RI/RO
structure to "bind" the ring rather than simply running a single line from
Port "1" on the switch to Port "1" on the MAU.

The ports of the switch plug into the MAU's just like any other node
on the ring. The RI/RO are used for connecting  multiple MAU's into a
ring.

First try to figure out how much bandwidth you're really going to need. It's
extremely difficult to tax a 16Mb token ring LAN, even with 130 machines. We
hardly ever see bandwidth limitation problems on local token rings with over
200 users. We have approximately 1300 individual token rings with anywhere
between 8 to 200 or so users per ring. The bandwidth limitations generally
form when a ring is bridged to another ring, and then only when it's a
remote bridge, local bridges can approach ring speed and are largely viewed
to be transparent from a bandwidth perspective. I would be very surprised to
hear you have maxed out on a 16 Mb ring with 130 users. You're likely to run
into other problems before bandwidth really becomes your problem.
Congestion, faulty wiring, etc. etc. are likely to be your problems before
bandwidth.

When you decide to move to a multiple ring configuration, whether through a
bridge or through a switch, you break down the single ring into multiple
segments and provide for some type of connectivity. Bridges connect two
segments together. Routers do the same thing but with more intelligence.
Each segment has one token to service it's clients with. Switches provide
for a token per port. We are moving to high speed switches only in our large
buildings with over 1000 users where there is already a local 16 Mb backbone
in place and 16 Mb locally bridged user rings. The switch does provide for
more robust, speedier service, but you may be using an elephant gun to kill
a mouse.

If you're even able to run at 4 Mb, the jump to 16 will be major, in terms
of capacity, and if you're not really taxing the 4, you won't even notice an
increase in performance. You have to be hitting the wall at 4 regularly to
be able to enjoy the additional bandwidth of a 16 and notice it.

I'd recommend the Olicom OC-8600 or OC-8601 series.  Rock solid
reliable, easy to configure (if you have to), fast, and HSTR capable.
We replaced a bunch of IBM gear with these, and couldn't be happier.
Reasonably priced too.

As far as connection goes, the OC-8600 is a 20 port switch.  Connect
one 8228 to each port on the OC-8600, and you are good to go.
Everything else is pretty much automatic, but you might need to set
the ring speed on the ports (easy to do with the supplied utility).
Just disconnect or wrap the RI/RO ports on each box.  Voila!  A bunch
of little rings!

If you have ports left over on the switch, connect your servers to
them at full duplex (if your NICs support that).

It's really easier than it sounds.  If you are still confused, drop a
line.

Chad

On Sun, 23 May 1999 20:26:57 -0500, "Rob Thompson" <a @aaa.com>
wrote: