|Thread Links||Date Links|
|Thread Prev||Thread Next||Thread Index||Date Prev||Date Next||Date Index|
Carlos: Yes you are right that topology based scheduling, will help.
Also there will tend to be grouping of subscribers in a dense urban case.
However in a previous life, I did see PONs proposals that had significant numbers of
subscribers per PON with a hierarchical splitter structure. You may have only been
500m from your nearest splitter, but the one beyond that - the distribution splitter -
could have been a significant distance away.
I'm not sure if the reflective case works with
a hierarchy of splitters. Once again previous examples I've seen assumed upstream most of the
power got to the head end due to directional splitters, and laser power at the subscriber unit were
kept to a minimum due to cost of device and cooling, reliability issues, and amount of power needed
to be sent out to the CLE, as it needed to be powered by the utility to meet lifeline requirements.
No one wanted to bet their legal team against a backup battery for lifeline.
It was a long time ago though, and there's been a lot of packets under the bridge since the;-(
From: carlosal@xxxxxxxxxxxxxxxxxx [mailto:carlosal@xxxxxxxxxxxxxxxxxx]
Sent: Friday, November 23, 2001 12:21 PM
To: Crick, Bill [CAR:1A00:EXCH]
Cc: 'Horne, David M'; 'Angeloni Cesare, IT';
Subject: Re: [EFM-P2MP] RE: EDMA R-PON
I think that the performance of the entire system would depend on two
- the actual physical topology of the network, as you've pointed out.
- the order of transmission.
Thinking as a carrier, I believe that it does make sense to deploy
splitters closer to the customer. If I was to put splitters at the core of
the network, I believe that the much simpler P2P Ethernet is a better
choice. So it minimizes the problem #1.
Also, thinking about *real* networks, it highly probable that we'll have
clusters of customers relatively close together. By cleverly ordering the
transmission sequence, one can minimize the delay. I don't have any actual
data to back this, but it's actually pretty intuitive, and fairly easy to
check with some simulations.