The IP over ATM Mailing List Archive by date

[Gary Hanson <gary@kentrox.com>]

On Wed, 18 Oct 1995, George Swallow wrote:

> >i don't remember how AAL3/4 implements MID allocation, but vci allocation
> >protocol could be something similar.
> 
> I've only seen the scheme used for SMDS/CBDS.  This assumes taht the
> links are all point to point.  The MIDs serve no purpose than to
> distinguish the cells of one packet from the next, so the assignment
> of MIDs can be arbitrary. The destination address needed is in the
> packet itself.  It is this address that the SMDS server uses for its
> routing decision.  Once routed the rest of the cells of that packet
> (i.e. those with the same MID) are routed to the same VPI/VCI/MID.
> 
> The sending end just uses MIDs as it likes.  All cells on a MID must
> go to the same dstination.  When a MID ceases to be used, it may be
> reused for a different destination after some timeout period.

Just to avoid possible confusion, I would like to point out that, as with
the VPI and VCI values, a MID value is only of local significance on the
link on which it is used, and has no end-to-end significance whatsoever.

> This mechanism doesn't solve any of the problems of associating a
> unique ID with an endpoint of a multicast call.  I'm afraid we'll have
> to invent something.

I am not quite sure that I see the insurmountable problem(s) posed by
overlaid P-MP connections or its multicast server alternative.  Could
someone please re-iterate the key points for me.  Thanks.

It seems to me that for any new invention to solve the problem in such
a way that it scales beyond 255 or 1023 multicast groups, it would have
to include one or more of the following heretical concepts:
  1. Larger cells (with larger cell header fields)
  2. A new AAL with capabilities for distinguishing more flows
  3. A possibility of frequent collisions during reassembly

Concept #1 would steer us away from ATM as we know it.  Concept #2 would
be interesting to watch get fought over, safely from the sidelines only.
And concept #3 would certainly make it interesting to distinguish faults
due to cell loss/duplication/trashing from faults due to collisions.

Or is scaling not an issue of interest here?

Regards,
Gary


_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/
_/  ______                                                            _/
_/ | ///  |       TM   Gary Hanson                gary@kentrox.com    _/
_/ |   ADC|Kentrox     14375 NW Science Park Dr.  503-641-3321 (FAX)  _/
_/ |______|            Portland, Oregon  97229    800-733-5511 x6333  _/
_/                                                                    _/
_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/