🚀 go-pugleaf

In article <1992Sep8.224744.18743@tamsun.tamu.edu>, gagrawal@cs.tamu.edu (Gopal Agrawal) writes:
> While going through the data books of the AMD and National Semiconductor
> chip sets for FDDI, I noticed that while AMD provides a register (TSYNC) 
> in its MAC chip (for SAS) to represent the synchronous bandwidth, the N.S. 
> chips don't. They just say that it is up to the user to assure that the 
> synchronous frames queue does not take longer than the allocated bandwidth 
> to transmit. Does the latest version of SMT say where exactly the 
> synchronous bandwidth is to be handled. I mean, are there any registers 
> or memory locations in the SMT or MAC or any other layer which are responsible 
> to time the transmission of the synchronous frames ? 
> 
> How do the stations agree on the distribution of the bandwidth amongst 
> themselves ?


>From SMT 7.2, Section 8.7, starting on page 137

  This section describes the Synchronous Bandwidth Allocation (SBA) protocol 
  employed on an operational FDDI logical ring.  ...

  ... The protocol performs the allocation functions through a
  request-response frame exchange between a station using Synchronous
  Bandwidth and a Synchronous Bandwidth management process...

  ... The Synchronous Bandwidth management process is not defined as part
  of this protocol. ...

>From Section 8.7.2.1

  ... A station that wants to change its current allocation of Synchronous
  Bandwidth must request the new allocation amount from the Synchronous
  Bandwidth management process. ...

>From Section 8.7.3

  ... Because the SBA protocol uses as central allocator, it may be
  desirable to have backup allocators available to assure availability
  fo the service.  If more than one allocator exists, a protocol
  must be defined among the allocators to determine which station
  contains the active management process and which stations become 
  stand-by allocators. ...


In other words, the draft ANSI standard for FDDI Synchronous Bandwidth
is at most a sketch for a single vendor, proprietary network, because
the SBA management process is not defined.  (There are other, harder
technical holes mentioned in the full SMT 7.2 section 8.7).  That
synchronous bandwidth is finally as much as that is an amazing
achievement.  That Bob &co. were able to prevent the normal Standards
Process 5 years of wrangling to produce an overly complete and
significantly wrong specification is an incredible, impressive feat.
(I mean that!)  Or maybe it's just that after ~10 years, X3T9.5 is
getting tired.


Vernon Schryver,  vjs@sgi.com