[comp.dcom.sys.cisco] FDDI Dual-Homing SAS Attachments

jnd@sms.com (Jose Nabielsky) (11/16/90)

Do cisco FDDI Routers operate correctly as *dual-homed* SAS stations?
(Some people call this in "cleave" mode.)  (For "Why bother?," read on.)

We are dual-homing the FDDI routers to DAS concentrators, as follows:

	1.	The PHY_B interface connected to a PHY_M port off 
		a PRIMARY DAS concentrator.

	2.	The PHY_A interface connected to a PHY_M port off
		a SECONDARY DAS concentrator.  (This provides path redundancy.)

The behavior we *expect* from the dual-homed interface is as follows:

	1.	The PHY_B interface provides the default attachment to the
		FDDI backbone (through the PRIMARY DAS Concentrator).

	2.	The PHY_A interface provides a hot standby attachment
		(through the SECONDARY DAS Concentrator).
		
	3.	Upon detecting a hardware- or software-induced PHY_B
		disconnect, PHY_A automatically provides restoral attachment
		(through the SECONDARY DAS Concentrator).

	4.	When path continuity is restored to the PHY_B interface,
		the router reverts to its default PHY_B attachment.

(If you have not figured out this yet, our motivation is to avoid an
informal, physical-ring wiring topology, and implement instead a formal,
physical-star wiring topology on this FDDI Backbone.)

Not all routers need be dual-homed.  Those that are not, simply will
attach to the PRIMARY DAS Concentrator through the PHY_B interface, and
will shutdown the PHY_A interface (through the command interface).

Has any of you implemented the dual-homed attachment scheme with (or without)
cisco routers?  Does this scheme work TODAY under release 8.1(xx)? Any gotchas?

-Jose
 jnd@sms.com

my@dtg.nsc.com (Michael Yip) (11/16/90)

I think that when a DAS connect any off its two (A/B) Ports to
a concentrator(s), the B Port will be the main one while the
A Port will be a hot stand by.  Meaning that when both Ports
are physically connected to the concentrator and only one is
Active (use to transfer frames).  If the B Port is unable to 
connect for any reason, the A Port will try to connect.

Any FDDI Dual Attached Station should do that, not just a CISCO
FDDI router.  I think that it is in the FDDI ANSI spec, in the
PCM section of CMT section of SMT document.

I am not sure what Jose's question is since he basically just
stated the situation.  ;)

-- Mike
   my@dtg.nsc.com

merike@helix.nih.gov (Merike Kaeo) (11/16/90)

The FDDI ANSI spec does indeed specify that the thru mode should be
disabled and PHY B take precedence in a dual-homing situation.  However,
whether this is implemented correctly by the vendors is a valid concern.
I have heard rumors that as of now, no router vendor is implementing this 
properly but that this issue is being addressed.  I haven't had a chance
to test it out personally with our cisco routers because of lack of
concentrators and haven't heard yet from cisco directly.

Merike Kaeo

jnd@sms.com (Jose Nabielsky) (11/17/90)

The following is the reply I got from Robert Fletcher at cisco in reference
to my earlier "FDDI Dual-Homing SAS Attachments" message. (cisco has done it
right, again.)

-Jose
 jnd@sms.com

------------------------- START OF FORWARDED MESSAGE ---------------------

>> The behavior we *expect* from the dual-homed interface is as follows:
>> 
>> 	1.	The PHY_B interface provides the default attachment to the
>> 		FDDI backbone (through the PRIMARY DAS Concentrator).
>> 
>> 	2.	The PHY_A interface provides a hot standby attachment
>> 		(through the SECONDARY DAS Concentrator).
>> 		
>> 	3.	Upon detecting a hardware- or software-induced PHY_B
>> 		disconnect, PHY_A automatically provides restoral attachment
>> 		(through the SECONDARY DAS Concentrator).
>> 
>> 	4.	When path continuity is restored to the PHY_B interface,
>> 		the router reverts to its default PHY_B attachment.

	We do not give a bias to PHY_B, we go WRAP on the PHY that 
	completes PCM/CMT first. The other PHY will go into "stand-by"
	"sleep" mode.

	If PHY_B completes PCM/CMT first we will insert our MAC as wrap_b.
	If PHY_A completes PCM/CMT first we will insert our MAC as wrap_a.

	If the fiber is disconnect we revert to the secondary PHY (assuming
        it is connected). If the primary is re-connected, we DO NOT break
	the ring to re-install the primary.

	Either of the PHYs can be shutdown via an exec command.

	This works at 8.1(25), will work better at the next maintenance
	release of 8.1, and is incorporated in the next major release 8.2.

Fletcher 
cisco Systems

--------------------------- END OF FORWARDED MESSAGE ---------------------