🚀 go-pugleaf

In article <92283.160905DART@SLACVM.SLAC.STANFORD.EDU>, DART@SLACVM.SLAC.STANFORD.EDU (415) writes:
> ...
> Does one connect a DAS station as per FIG.1 or FIG. 2 below.  Figure 1 shows
> a  DAS station plugged directly into two concentrators.  Figure 2 shows
> the concentrators cross-coupled.  I believe, and I will be testing this,
> that FIG. 2 is the correct configuration as it's the only way I can see
> that normal token rotation will be maintained.   Does anyone have Dual
> Homing in place?  (My original note follows the figures.)
> 
>           |----------|            |----------|
>  TO     ->|A  CON    |->->->->->->|A         |
>         <-|          |<-<-<-<-<-<-|  SERVER  |
>  DUAL     |    1     |            |    1     |
>         ->|B         |       -->->|B         |
>  RING   <-|   DAS    |      /-<-<-|   DAS    |
>           |----------|     //     |----------|
>                           //
>           |----------|   //
>  TO     ->|A  CON    |-> /
>         <-|          |<-<
>  DUAL     |    1     |            FIG. 1
>         ->|B         |            Direct Dual Homing.
>  RING   <-|   DAS    |
>           |----------|
> 
>           |----------|            |----------|
>  TO     ->|A  CON    |->->->->->->|A         |
>         <-|          |<--       <-|  SERVER  |
>  DUAL     |    1     |   \     /  |    1     |
>         ->|B         |    <-<-<-<-|B         |
>  RING   <-|   DAS    |      /-<-<-|   DAS    |
>           |----------|     //     |----------|
>                           //
>           |----------|   //
>  TO     ->|A  CON    |<- /
>         <-|          |-->
>  DUAL     |    1     |            FIG. 2
>         ->|B         |            Cross-Coupled Dual Homing.
>  RING   <-|   DAS    |
>           |----------|


Fig. 1 is correct.  Fig. 2 will not work with ANSI standard
concentrators or stations.  You can't split connections that way.
CMT signalling will not work in Fig. 2.

All SMT 7.2 compliant DAS vendors "have dual homing in place"
(including Silicon Graphics, of course.)  Dual homing was in use in the
FDDI demo ring at Interop91.

"Dual homing" is only a scheme for having a "hot backup".  When the
connection to the B port of the server fails, the server will finish
CMT signaling on the A port and start using the A port.  When the B
port looks good again, as determined by still more CMT signalling, the
server is required by the SMT standard to shut down the A port and
start using the B port again.

I've grumbled that this is not good for moving customer packets (e.g.
B might pass LCT but not be able to move data--I've seen real cases of
that), but people in X3T9.5 have told me that the more consistent ring
maps produced by the standard mandated behavior is more important.


Vernon Schryver,  vjs@sgi.com