🚀 go-pugleaf

vjs@rhyolite.wpd.sgi.com (Vernon Schryver) writes:

>In article <1992Oct12.160456.11141@walter.cray.com>, hens@snipe.cray.com (Tom Stephens) writes:
>> 
>> Can someone tell me how they really work?  Or better yet, point me in
>> a direction where I can read about them?  I am having problems
>> explaining that if two machines are powered off in an fddi ring, they
>> will not isolate the machine in the middle.  Senerio..
>> ...


>Think of tiny mirrors controlled by solenoids with springs.  Literally.

>Apply power (3-5 v, I've forgotten how many ma but little) to the 2
>"signal" wires to the switch, and the mirrors passively direct photons
>from the neighbors to the station.  Remove the power, and the mirrors
>move back to the "de-inserted" positions, passively route photons
>from the neighbors around the powered off or just de-inserted machine.

>Very elegant and reasonably reliable if you don't mind the quirks of
>mechanical stuff (vibration, shock, and orientation susptibility,
>"stickiness", loud click when changing state.  Never mind what their
>makers claim; those seem to be large or small in all of them.)


>Vernon Schryver,  vjs@sgi.com

Actually I think the actuators are bi-metallic devices that heat and bend
when the control leads are activated (+5VDC I seem to recall) The mirrors
are usually prisms I think. Here's a shot at a crude ACSII diagram:


             Bypass not activated                     Bypass Activated

Backbone side


                A            B                         A            B
              o   o        o   o                     o   o        o   o
              :   :        :   :                     :   :        :   :
              :   :        :   :                     :   :        :   :
              :  |:        :|  :                     :   \--------/   :
              :   :        :   :                     :                :
             |:   :        :   :|                    \----------------/
              :   :        :   :                                       
              :   :        :   :                                       
              o   o        o   o                     o   o        o   o
                A     @      B                         A      @     B
                      |                                       |
Device side           |                                       |
                      |                                       |
                      |                                       |
                control lead                            control lead
                supplying +5VDC                         supplying 0VDC
                         


The way the control leads work is that while the attached device is
powered on +5VDC is provided to the bypass switch which is used to
swing the mirrored devices out of the optical path. When the attached
device is powered off (or SMT switches off the optical bypass) the
+5VDC is removed from the control leads and the mirrors/prisms swing
back into the optical path thus switching the light back out of the
bypass switch.

The issue with using bypass switches is power budget. Specifically
how many optical bypasses can you traverse if several of them *IN A ROW*
are tripped. For example

    A--B--c--d--e--f--g--h--i--j--k--l--m--n--O--P

The letters above represent devices on a portion of a ring each
with an optical bypass switch attached. The capital letters are
active devices and the lower case letters are devices that have
their bypass switch engaged. In this example the question is:

"Is there sufficient light (optical power) sourced from B and O to
overcome the attenuation introduced by traversing the connectors
of bypass switches c--d--e--f--g--h--i--j--k--l--m--n ?"

Is it realistic for several consecutive neighbors to go down at the same
time? What about a whole building loosing power in a campus environment?
These question are just food for thought.

Personally I don't have a strong opinion one way or the other.

-brad-

--
v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v^v
Brad Turner | 5400 Bayfront Plaza  | Marketing Engineer   | (408) 764-5261
3Com Corp.  | Santa Clara CA, 95052| mbt@NSD.3Com.Com     | (408) 764-5002 fax