| Hi,
Already thanks for your info's ;
The reason why the customer is ask me that is that he wants to be
able to short circuit a bridge in case of failure.
When you say that it is an active device, does-it means that the
by-pass is to be connectected in parallel with the DAS bridge
and will automatically loop-back each side or does-it need a manual
intervention .
Is it possible to be a little more precise when saying that
this is inefficient, so that I can persuade him to not use it.
Again the customer is expecting the price of a by-pass in the offer
so do you think that it should be available by DEC.
Regards,
Robert
|
| Re .1, .2, you should tone down those comments a bit. Even if we didn't
offer bypass relays, you're overdoing it -- but in fact we do offer them.
Bypass relays are used with dual attached stations (or with
dual attached concentrators; I don't know if we offer that yet). The
two fiber pairs from the two neighbors are plugged into the relay, and
the other two ports of the relay plug into the DAS A/B ports. The relay
has two positions: inserted, where one input is fed through to the A and
the other to the B port of the station -- and bypassed, where the two
inputs are connected to each other and not to the station. The bypassed
state is the default, in other words that's what you get when no power is
supplied.
Bypass relays are electromechanical devices. They have moving parts, but
no electronics. They also have signal loss.
A bypass relay will bypass a station that is turned off, or has had its
power removed. Of course it does NOT protect from other problems, such
as defective station hardware that leaves the station on but working
incorrectly; unplugging of the station (with its relay) from the fiber
cables; etc.
The major difficulty with bypass relays is that they introduce optical
signal loss. If you compute path losses from the "worst case" numbers
used in typical cable plant design procedures, you'll discover that you
can SOMETIMES afford a single bypassed station in the path, but it's highly
unlikely you could have more than one bypassed station in a row and still
have enough signal. In MANY cases, even a single bypassed station will
cause you to have excessive loss.
The result of excessive loss is excessive bit error rates. What can happen,
depending on the exact numbers, is that there is enough signal for the
connection to remain active, but the token and frame loss rate is quite high.
This means (a) your network performance is bad, (b) you have a much higher
probability of UNDETECTED data corruption. (This is because the CRC catches
ALMOST all, but not quite all, data errors. The higher the bit error rate,
the more often you play the CRC lottery...)
If you actually measure signal losses in an installed cable plant, you may
discover that the losses are noticeably lower than the design rules suggest.
That's because the design rules allow for cable variations, connector quality
variations, and component aging. If you use measured loss numbers, you might
decide that you can afford to use bypass relays even when the design rules
say you cannot. This is risky, since you may find that 6 months later the
loss numbers aren't as good anymore and your network now no longer works
when there's a bypass.
What does the customer really want? I expect they want redundancy: the
network should remain fully available even if one (or even more than one)
station is down.
There are a number of ways to achieve that. Bypass relays provide one way,
but with significant limitations and cautions. A small dual ring built
from concentrators, with DAS stations attached to that with dual homing,
will give you the same or better redundancy without the loss and distance
issues. Finally, if you're talking about bridges or routers, you can
simply add redundant bridges or routers, making use of the higher layer
facilities (bridge spanning tree, routers' dynamic routing algorithms) to
provide far MORE redundancy than either of the two low-level mechanisms
would give you. This does involve additional boxes, though, so this extra
capability does come at some cost. Often that cost is quite acceptable, though.
If the customer has absolutely decided to use bypass relays, we can do that.
Tell them that bypass relays have significant limitations and we generally
recommend other alternatives, but we do offer them if that's what he wants.
paul
|
|
OK My .2 reply did not explain in detail my point.
I'm used to talking to other technical people and when trying to
explain something to someone non-technical I don't always come across
properly.
My point was the same as -.1 in that they are mechanical devices and
add a really crappy loss into the network. In the DVN broadcasts on
FDDI they were downplayed by our own people as not really a viable
alternative as of this time. With improvements in technology and
optical switching they may be ok in the future.
If a DAS device goes out, the ring is going to loop anyway and you will
still have a working network (Dual Ring ???). The devices below that
leg of the DAS device will of course no longer be connected to the main
ring, but should continue to communicate within that tree structure.
Only if 2 DAS devices go out will you then segment off into 2 seperate
network segments. Redundancy is always a good idea, but can be taken to
extremes. Our FDDI products seem very robust in design, and after
introducing multiple faults into a 6 concentrator Dual Ring, I am very
impressed with the intelligence of the technology.
MTS
|
| Well, so long as two consecutive stations never bypass at the same time,
it might work.
We can sell what's being asked for, and if the customer absolutely insists
on not changing his mind, we should. First you should try to change his
mind, though. Use the comments in 320.4.
paul
|