Conference 7.286::fddi

1842.0. "Failover DECswitch900EF 2x ???" by 54690::KUPPENS (I know nothing, I'm from Barcelona...) Thu Oct 26 1995 11:28

    

Trying to build a failover scenario for a OEM customer:


			Rest of network
			---------------------

	|-----------------------|	|---------------------|
	|DEChub 900		|	|DEChub 900	      |	
	|			|	|		      |
	|			| FDDI	|		      |
	|DECswitch 900 EF------------------DECswitch 900 EF   |
	|Portswitch 900 16x BNC	|	|Portswitch 900 16x BNC|
	------------------------|	|---------------------|
		     \			       /
		 16x   \		      /  16x
			\MUX1-MUX2-MUX3-MUX4/
			\MUX5-MUX6-MUX7-MUX8/
			\MUX9-MUX-MUX-MUX12/
			\MUX13-MUX-MUX-MUX16/
			\MUX-MUX-MUX-MUX/
				.
				.
			\MUX-MUX-MUX-MUX/
			\MUX-MUX-MUX-MUX64/


			Customer's MUXen (running OS9)





The customer's MUXen are OS9 broadcast-boxes (BNC-based) which have to be up
and running  24 hours per day (supplying broadcast streams).


What will happen if the DECswitch 900 EF  fails ???,  how long will it take
to have the  2nd  DECswitch 900 EF take over completely, so that the MUX-en
can use the 2nd DECswitch (via PORTswitch 900) ?. Is the DECswitch 900 EF
a "real" switchbox (as it used to be called DECbridge 900 we have our
doubts ????), because if it is the scenario described above implies that the
DECswitch is actually a number of bridges, and that will imply that with the
Spanning Tree algorithm it will take a lot of time to have the
2nd DECswitch 900 EF build up the new table (how much ???).
If however the DECswitch 900 EF *is* a real switch we expect the
2nd DECswitch 900 EF to be up and configured with all network info
within 40 seconds (Spanning Tree specs ?).


With 3COMs  LANPLEX 2016's a takes approx.  45 sec. per port, and as it
supplies 16 BNC ports (no need for seperate PORTswitch-like box to supply
needed BNC) in total it takes 16x 45 sec. to failover to a second
LANPLEX 2016, which is not really an acceptable failover (need a scenario
which fails over within a minute or faster!). We actually tried this,
and best we can get is to lower forward delay, and thus getting a max
switchover time of 5 minutes, which is still unacceptable.


As the Gigaswitch is way out of reach for price-issues (and cannot
wait for DECswitch 400 or other possible to be delivered near future
products) we hope that the DECswitch 900 EF  *is*  a real switch and
that it can confirmed that it will take over the functionality of
the first switch within 1 minute ????


ThanXs,
Robbert.


    
P.S.:  If this is not the appropriate notesfile, which is ?
    
    
    

    
    According to Karl Pieper (PM) the DECswitch 900 is a multiport bridge,
    so failover should be possible within 45 seconds for the entire switch,
    
    --Robbert.