| Your topology would recover, but it is not optimal. In your
diagram, the most significant recovery times are all spanning tree
related. In other words, there's nothing special to the
GIGAswitch/FDDI system - it depends on the spanning tree parameters.
By default a failure involving the change of root port would be the
typical 50 seconds for root related reconfiguration.
Most people would use a simple dual homing topology with two
GIGAswitch/FDDI systems for redundancy.
See the following notes for previous discussion on this topic:
Notes> show keyword *failover-config*/full
GIGAswitch
Created: 29-JUL-1992 12:28 923 topics Updated: 11-FEB-1997 14:39
-< GIGAswitch/FDDI Jan 97 BL3.1 914.0 documentation 412.1 >-
Keyword Note
>FAILOVER-CONFIGURATI 175.18, 175.19, 175.22, 175.23, 363.7, 510.10, 517.0,
517.1, 527.8, 527.9, 527.10, 540.0, 540.1, 540.2, 540.3,
540.4, 540.5, 540.6, 580.11, 743.8, 743.9, 743.10,
743.11, 743.12, 743.26, 868.7
End of requested listing
|
| re:.1
Thanks for your input.
This config has until now consisted of one Giga with some servers
connected to FGL-4 ports and four FDDI campus-wide rings connected to
its FGL-2 ports.
Now they will increase the number of servers directly connected to the
Giga and also dual-home these. A second Giga will be placed besides the
first one with a hunt group between, just to get enough ports. There will
also be a hunt group to a third Giga in another building.
Our idea was that the four FDDI rings should be independent of the
loss of one of the Gigaswitches. Could this be accomplished in an
easier way maybe? What would happen if we just connected both Gigas as
DAS stations to all four rings, (keeping the path cost for the hunt
group between them lower than the path cost for the DAS ports to the
rings) for example?
Thanks,
Nils (working together with Roger).
|
| Connecting multiple GIGAswitch/FDDI systems to the same ring will
work, but it isn't the best use of materials. If you do this there is
no possibility of FFDT (full duplex) operation. Also, since you have
created loops, you are forcing yourself into spanning tree failover
scenarios.
The standard dual homing scenario works fine. Connect all the dual
homed stations B ports to a "B" GIGAswitch/FDDI system. Connect all
the dual homed stations A ports to a "A" GIGAswitch/FDDI system.
Connect the two GIGAswitch/FDDI systems with a hunt group.
If you feel that you can control the wiring changes completely and
avoid loops, you can turn spanning tree off on all the M ports and the
hunt group ports. This avoids some of the spanning tree failover
delays.
If you can't avoid loops, don't turn spanning tree off.
You can dual home the third GIGAswitch/FDDI system to the other two
GIGAswitch/FDDI systems, using DAS interfaces for the hunt group
connections, and dual homing the hunt group ports to the two different
GIGAswitch/FDDI systems. I.E. dual homing the hunt group.
For a dual homed hunt group with two members, if one member fails,
the hunt group will stay up as a single member on the original path
with diminished throughput (the single member).
(You may want to locate the "B" ports on the same card - that way
if the card dies, both ports will fail over and the hunt group will
become active on the "A" GIGAswitch/FDDI system, instead of running
with just one active hunt group member).
If both ports die, the hunt group port will go down on the "B"
GIGAswitch/FDDI system, and the hunt group port on the "A"
GIGAswitch/FDDI system will become active.
A three member dual homed hunt group operates similarly, but keep
in mind that if only two members die, the hunt group still won't fail
over until all the members are dead.
You can keep dual homing additional GIGAswitch/FDDI systems as
required (until you run out of ports).
MDL
|