Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sun-barr!decwrl!sgi!vjs@rhyolite.wpd.sgi.com
From: vjs@rhyolite.wpd.sgi.com (Vernon Schryver)
Newsgroups: comp.dcom.lans
Subject: Re: Does such a "box" exist?
Keywords: network star fddi concentrator
Message-ID: <86019@sgi.sgi.com>
Date: 19 Feb 91 00:28:15 GMT
References: <3517@uc.msc.umn.edu>
Sender: guest@sgi.sgi.com
Organization: Silicon Graphics, Inc., Mountain View, CA
Lines: 62


There are FDDI "concentrators" which connect one or more stations to the
dual ring.  Most FDDI concentrators should be thought of as complete
FDDI stations which also have electrical switches that connect additional,
"slave" optical cables in series with the concentrator's own hardware on
the dual ring.

Whether you use concentrators or not, it is good to connect FDDI machines
in a star, with all fibers connected to a patch panel.  A patch panel
allows easy, manual reconfiguration, and costs little more when expensive
FDDI electronics are considered.  All of the big fiber vendors make MIC
(FDDI) patch panels.

Ignoring patch panels and nonstandard topologies, there are 3 main ways to
connect FDDI stations:
    1. single or dual attach stations with concentrators.
    2. dual attach with optical bypass switches.
    3. dual attach without switches.

Method 1 is most robust, but most expensive with lower potential performance.
  Concentrators require more optical equipement and electronics per station
  than other solutions.  Optical bypass switches and the ODL (electrical-
  to-optical hardware) are currently the most expensive FDDI parts.
  A concentrator typically costs 0.25 more ODL's/station but saves 1.0 to
  0.75 optical bypasses/station.  (If you need robustness, you need
  bypasses on the A/B ports of your concentrators.  4-port concentrators
  seem to be popular.)

  Using concentrators limits you <100Mbit/sec.  There is hope of using
  dual-MAC-dual-attach FDDI stations to get >100Mb.  (The birthday
  paradox ensures the 2nd ring is almost always available for data.)

Method 2 is almost as robust as 1, except that only a limited number of
  consecutive stations can be "bypassed" with switches.  (When the power
  goes off or the station is reset, the switch passively drops into the
  "bypassed" state.)  With short cable runs and newer switches, this is
  often not a problem.  Switches are not cheap; they're mechanical devices.

Method 3 is fine in a computer room where the operators can run over
  to the patch panel when the mainframe field engineers want to take down a
  system.  It is problematic elsewhere.


Most people and all FDDI concentrator vendors and central computer facility
managers say that concentrators are the only way to go.  They note
concentrators allow multiple stations to be powered off without breaking
the ring.  (Since FDDI is a dual-ring, any single station can be powered
off without breaking the ring, regardless of concentrators or bypasses.)

A surprising number of people do not know what a concentrator does.  They
claim an FDDI concentrator can somehow isolate the ring from a misbehaving
station.  This is false.  All FDDI frames pass through all active FDDI
stations.  The ring is not in any way protected from bogus frames by
concentrators.  A concentrator can isolate stations that cannot do "CMT
signalling," but famous trade shows have demonstrated that is not a
problem.  The infamous trade show problems have been frame problems, and
were not affected by concentrators, or, in one case I know of, were caused
by bugs in the concentrators themselves.



Vernon Schryver,  vjs@sgi.com