Conference 7.286::fddi

	No.  Can't do that.

	You can NOT have a Peer connection AND a Tree connection on your
	concentrator's AB pair.

	Tree connection = anytime you have an M connection 
	Peer connection = anytime you don't have Ms

	So A to M is a Tree connection.  B to A is a Peer connection.

	A 'dual-homed' concentrator has an A to M and a B to M connection.
	That is 2 Tree connections which is OK.

	A concentrator on the truck ring has an A to B and a B to A connection.
	This is 2 Peer connections which is also OK.

	I'm not sure which connections you are trying to protect but I'll
	assume you want the GIGAswitch.  So in that case I would connect
	my three concentrators on the trunk ring (A to B, B to A) and then
	connect my switch A port to an M on one concentrator and the B port
	to an M on another concentrator.

	Does that help,
	Luc

>	I'm not sure which connections you are trying to protect but I'll
>	assume you want the GIGAswitch.  So in that case I would connect
>	my three concentrators on the trunk ring (A to B, B to A) and then
>	connect my switch A port to an M on one concentrator and the B port
>	to an M on another concentrator.

    The intent of the configuration seems to be to bridge three rings
    through GIGAswitch when the DECcon-to-GIGAsw links are working,
    and to concentrate the three rings into one with the GIGAsw is down.
    So the proposal above is not a solution.  The text above never
    bridges between 3 rings; it always has 1 ring, and wastes 1 GIGAsw
    port when the GIGAsw is operational.

    Since GIGAsw has no M ports, there is no way to have a concentrator
    talk to GIGAsw when GIGAsw is up, and talk to another DECcon when the
    GIGAsw is down.  At least, there is no way to do it that obeys the
    rule that dual-homed concentrators need either two tree connections
    or two peer connections.

    That was a concentrator rule.  What rule applies to end stations (and
    bridges and routers)?  Is it also impossible for a DAS station to have
    port B connected to an S port on GIGAsw, and port A connected to an
    M port on a concentrator?  I assume so.  So DAS stations must either
    be dual-homed to two concentrators, or to two S ports on (one or) two
    FDDI bridges (e.g., GIGAsw).

    Sadly, there is no way for a DECconcentrator to offer backup connectivity
    for a GIGAswitch.  If GIGAsw had M ports instead of S ports, would that
    solve this problem?  Or would SMT go nuts because a DAS device is "dual
    homed" to two completely unrelated FDDI rings (the bridge ring and the
    backup concentrator ring)?

    --gw (who never read the FDDI config rules, CMT or SMT)

Yes, if a Gigaswitch had M ports, you could dual home to it.

No, SMT wouldn't know that you're dual homing onto separate LANs.  For that
matter, that can occur anyway -- if you have a twisted dual ring, for example,
so some concentrators are bringing down the secondary ring and some the
primary.  Or if you have configured the concentrator to attach some of the M
ports to the secondary path.  Neither of those are things you can do with
DEC concentrators, but they are permitted by the standard and some other
concentrators do implement it.  (The SMT standard is designed on the theory
that customers must be given as much rope as possible to hang themselves with.)

So it's already the case that a dual homing failover could cause some confusion
because you're moved to another LAN.  An M-ported Gigaswitch would simply
create another instance of this.  Given that in the Gigaswitch case the two
LANs in question are bridged together, the actual impact on any higher layer
protocol is almost certainly zero.

	paul

  I guess it should be stated here for completeness that the when the four port
SAS interface (S ports only) for the GIGAsw becomes available it will not be
possible to connect either an A or B port to this device.

  But it is possible to connect the M port on the concentrator to the S port
on the GIGAsw or visa-versa.

Huh?  A to S and B to S are legal connections.

	paul

Based on the rules below, and with the intent of providing for a "switch"
based primary configuration ( for performance ), and a "tree" fallback
configuration:

	1)  Must all dual homing occur through an A-M/B-M connection
		( A-M/B-S will not dual home)?
	2)  Since the chart below indicates it is legal, what happens instead
		of dual homing with an A-M/B-S connection?

Thanks,

Richard Robak

--------------------------------------------------------------------------------

The following is text taken from "DECconcentrator 500 Installation" ( Digital
Part# EK-DEFCN-IN-002 ) Page 4-6:

Table 4-1: Connection Rules

----------------------------------------------------------------------------
Connecting Nodes   This Node       This Node       This Node       This Node
----------------------------------------------------------------------------
Port Type          A               B               M               S
----------------------------------------------------------------------------
Other Node  A      No              Yes, rule 1(1)  Yes             Yes
Other Node  B      Yes, rule 2(2)  No              Yes             Yes
Other Node  M      Yes, rule 3(3)  Yes, rule 4(4)  No              Yes
Other Node  S      Yes             Yes             Yes             Yes
----------------------------------------------------------------------------
1 Connection Rule 1: This Node Port B can connect to Port A of Other Node
  if, and only if, This Node Port A is connected to Port B or Port S of
  another station, or is unconnected.
2 Connection Rule 2: This Node Port A can connect to Port B of Other Node
  if, and only if, This Node Port B is connected to Port A or Port S of
  another station, or is unconnected.
3 Connection Rule 3: This Node Port A can connect to Port M of Other Node
  if, and only if, This Node Port B is connected to Port A or Port S of
  another station, or is unconnected. If This Node Port B is connected
  to Port A of another station, it immediately disconnects when This
  Node Port A connects to Port M of Other Node.
4 Connection Rule 4: This Node Port B can accept connections to Port M of
  Other Node. If This Node Port A is connected to Port B or Port M of
  another station, it immediately disconnects when This Node Port B
  connects to Port M of Other Node.

As the chart will tell you, connecting A-M and B-S will result in both
connections coming up.  In effect you're using the B port as if it were an M
port.

	paul

    Connection Policy:

	First of all those are the DEC connection rules which will not
	apply to products that run SMT 7.2.  These new products will match
	the connection rules of SMT 7.2 (good and bad).  Most of the rules
	are the same but I just want to clear that up.

    CFM:

	As far as the Peer/Tree connections, the Tree connection will
	come up but the Peer will not.  This is not what SMT says.  SMT
	says B will come up.  We (DEC) feel that it is better to have
	the Peer connection (A or B) come up instead.

    Three more questions...
    
    First ( and most importantly ):
    
    	1) Will M port boards for the Gigaswitch ever become available, and
    	   if so when?
    
    		Although the customer is enticed by the performance gains
    		offered by a Gigaswitch, he is VERY sensitive about network
    		availability and will not accept a single point of failure.
    		I have talked to him about Gigaswitch reliability, but he
    		refuses to purchase the Gigaswitch unless it has a way to
    		be "backed up."  Is there some other way this can be
    		configured so he can profit from the performance of the
    		Gigaswitch and still meet his demands for network
    		availability if the Gigaswitch dies?
    
    
    Secondly:
    
    	2) How does the Gigaswitch decide how to configure its ports?  If
    	   you plug in two PMDs on one port do they automatically configure
    	   as A/B ports or is some other mechanism involved in this ( if so,
    	   what is it? )?  If you have one PMD do you plug it into the A or
    	   B (top or bottom) PMD slot?  Once it's plugged in does it
    	   automatically configure itself as an S port or do you have the
    	   option of configuring it as an A, B or S port?
    
    
    And last:
    
    	3) Based on the rules listed in .6, the configuration in .0 is
    	   legal - but what really happens when all the ports become active?
    	   Does the token actually pass through the Gigaswitch multiple
    	   times, and if so how does the Gigaswitch (and network) deal with
    	   this?  Conceptually, I have a hard time visualizing the path
    	   configured by .0.
    
    
    Thanks,
    
    Ricahrd Robak

	1-) I don't work on GIGAswitch so I can't answer that.  But
	it's all software...  That is, there is no hardware difference (except
	for keys) between an M port and a B or S or A.  The software
	sets up the configuration.  If GIGAswitch wants to become a
	2 port NAC (Null attach concentrator) it is possible.  But it
	would be mean more configuration info (MIBs).
	
	2-) 2 ports means DAS automatically.  I'm not sure about only one
	PMD.  I think if you plug it into the top it will be an S port.
	Otherwise it's a DAS with the bottom port being a B port (A port
	being unused).

	3-) No token on the 'unwanted' connection (Peer connection).
	It's almost as if the cable isn't there.  Just like a 'backup'
	link in dual homing.

	
	So, I think the only way to do what you want (stop the GIGAswitch)
	from being a single point of failure, is to have the GIGAswitch
	become a 'bridge/concentrator'.  That is a concentrator (NAC) that
	does bridging.  Sounds like a good idea to me.

	Luc