| You mean within a given fiber pair connecting two ports, the fiber one way is
a different length than the other?
That's odd but legal. Be careful to use the worst case total fiber length
when checking configuration rules, allowing for a wrap in the worst possible
place.
paul
|
|
> You mean within a given fiber pair connecting two ports, the fiber one
> way is a different length than the other?
Yes, 21km and 8km.
The customers proposed configuration includes 4 DECconcentrators (2 at each
site).
SITE1 SITE2
SMF
-------------------------------------
A - - - - - - - - - - - - - - - - - - - B
DC500 DC500
B A
| |
MMF | | MMF
| |
A B
DC500 DC500
B ------------------------------------- A
- - - - - - - - - - - - - - - - - - -
SMF
primary = ----- secondary = - - -
They will be running dark fiber between the two sites using two separate
suppliers. They plan on using one supplier for the primary ring and another
for the secondary. This infers that both primary intersite fibers would travel
the same physical path and both secondary intersite fibers would travel a
different physical path. I have some concerns with this. What happens if
both the primary fibers are severed or the communication is interrupted in
some way leaving only the secondary fibers intact?
The port would have to recognize that the primary was bad, the secondary was
okay and use the secondary. If the port were to wrap we would be in trouble as
there would be two separate disjoint rings.
I would rather see them use separate suppliers for each port to port link. In
this case if the link went down the port would wrap and all would be fine.
|
| This sort of approach is what "global hold" was invented for, but that's a very
marginal solution for a very limited class of stations. (Its originators don't
seem to like it much anymore either...)
FDDI was designed on the assumption that the fiber pair connecting two ports
would be routed together. That's not written anywhere, nor is it required; it's
simply the assumption that underlies much of how FDDI works.
The customer assumption seems to be "only the primary ring carries data so only
the primary fiber needs to be operational". That's not how FDDI works. If
either fiber in the port-to-port connection fails, that connection is shut down
and the ring wraps.
If only a single fiber breaks, you have one wrap. In that case it doesn't
matter much whether your fibers are all from the same company, or from two
companies.
If you worry about what happens when ALL the fibers supplied by one of your
companies fails, the answer is: "your network is down". If you built the
topology shown out of dual-MAC stations that support Global Hold (if there
are any such) you'd still have a running network. (If anything glitches,
your network would break, and it would NOT come back -- the reinitialization
requires both fibers to work whether global hold is supported or not!) With
"normal" FDDI stations, this scenario will give you wraps at every point,
i.e., no connectivity at all.
The right solution is to use a dual homed tree. Put the root in a single
place. (If you're worried about that place burning down, put half in one place,
half in another, with two fiber pairs, one from each company, connecting the
two sites.) Run redundant connections, one fiber pair from company A and
the other pair from company B, to each other point.
If you do this, you still have a working network even if every fiber supplied
by company A goes out of service simultaneously.
paul
PS. This is a very nice example to demonstrate the fact that dual homing gives
you MUCH MORE fault tolerance than the "traditional" FDDI techniques such as
a dual ring with wrap, or even the "highly fault tolerant Navy spec" kludge
called Global Hold.
|