Conference netcad::hub_mgnt

3116.0. "High Availability, Fault Tolerant Bridge Config ??" by MEOC02::MCCOY (want peace ? -> work for justice) Thu Jan 04 1996 00:17

We have a customer who requires a highly available network configuration for a
air traffic control application.  The following explanation and diagram is the
customer's proposed Fault Tolerant configuration.  They need to understand the
expected switch-over times would be in the event of a failure of any single
segment or device.  They want 10 seconds or less!
                            
Basically there are three logical LANs each with several devices attached to
them via fault tolerant transceivers (FTT in the diagram). The diagram does not
show the 900TM repeaters which physically connect the FTT's to the 900EF's. The
primary and secondary LAN segments are bridged together.  The three LANs are
then bridged together in a redundant configuration.  All segments are 10Mbps
ethernet and TCP/IP protocol is used exclusively.
                                                       
In normal operation assume that 900EE#2 is the root bridge, and segments
secondary A and secondary B connected at 900EE#5 are in standby state.  All
traffic between the LANs will flow via primary A and primary B.  

      LAN 1	              LAN 2		     	LAN 3
vvvvvvvvvvvvvvvvv      vvvvvvvvvvvvvvvvvvv      vvvvvvvvvvvvvvvvv

                 primary A                primary B
	900EE#1 --------------- 900EE#2 ---------------	900EE#3
       /   |		       /   |		       /   |
      /	   |		      /    |		      /    |
     /	   |		     /	   |		     /     |
 -----	   |		 -----	   |		 -----     |
| FTT |	   |link1	| FTT |	   |link2	| FTT |    |link3
 -----	   |		 -----	   |		 -----     |
     \	   |		     \     |		     \     |
      \	   |		      \	   |		      \    |
       \   |		       \   |		       \   |
	900EE#4 ---------------	900EE#5 ---------------	900EE#6
                 secondary A             secondary B

                            
Fault Scenario 1.
=================
In the event of a UTP cable failure between the FTT and 900EE#1, traffic will
be sent via 900EE#4.  Changeover time will be a matter of milli-seconds for the
FTT and a few seconds as the bridge tables in 900EE#1 and 900EE#4 are updated.
Q.  What is the expected bridge table learning time in this scenario ?

Fault Scenario 2.
=================
If 900EE#1 fails then 
	o	900EE#4 must learn the attached devices 
	o	the link secondary A must change from standby to active 
	o	forwarding tables in 900EE#5 and 900EE#2 must be updated

Q.  What are the pertinent times, and are there any tuneable parameters on the
900EE to minimise the transition time ? 

Fault Scenario 3. 
================= 
The worst possible scenario would be when 900EE#2 (the root bridge) fails and a
spanning tree configuration is required.  From experience, the default values
in the 900EE would see a spanning tree reconfiguration of about 45-50 seconds.  

Q.  How short can this be configured to, and what parameters need to be altered
on the 900EE from hubwatch.                   

General Question
================
Can anyone suggest an alternative topology which gives three separate LANs and
no single point of failure which would have an inherently lower transition
time in the event of any expected fault conditions ?