Re: [EFM] P2MP -> P2P migration path
At 09:48 17/07/01 -0700, John Pickens wrote:
>One school of thought is that EPON is just a simple transition technology
>on the way toward PTP, so (upstream) efficiency is no
>consideration. Heck, 10-20% of available link capacity would be
>great. Heck, why not just divide up the bandwidth statically and give
>each ONU 1/16th of the raw link bandwidth. It's a lot simpler that
>way. Plus the EFM operator will just switchover to PTP anyway once the
>traffic gets to a significant level.
>Another school of thought is that EPON is an enduring technology which (in
>order to compete with other enduring shared technologies, e.g. copper and
>vaporized-copper) must dynamically maximize the utilization of the link
>bandwidth up to 100% utilization. Furthermore latency sensitive traffic
>must be capable of seeing minimum latency (e.g. 0 ms for a constant bit
>rate stream) on the uplink.
>So the question is which extreme model (or intermediate model) is
>required. The model chosen will determine the functionality requirements
>of the MAC/PHY.
I could not have put it better. That's where my wavelength allocation
request comes into play. As we don't know (as of now) which model is going
to be the winner, I believe that we have to reserve space for the P2P
migration path. A carefully designed wavelength allocation scheme may allow
this discussion be be tackled later. For now, we can work out the P2MP EPON
standard, knowing that we will have room for P2P in the future.