said by buckweet1980:
In the FIOS use case the OLT devices are really nothing more than a bridge/transceiver to get the traffic off of the GPON medium and onto Ethernet at the CO. It doesn't really care about the layer-2 conversations, it just processes them transparently. The real work on FIOS is being done by the routers sitting behind the OLT boxes. FIOS for my node is serviced by a Juniper E series router, which I assume is common across most areas. The E series have been around for a long time and are specifically built for handling 10's of thousands of subscribers on a single box.
Not as transparent as what GPON vendors want buyers to believe. The OLT still has to maintain a very long list of MAC addressing table and ONT IDs. It also have to continuously identify/monitor for ONTs for bandwidth mapping, time all the slots accordingly and GEM packaging for the 10s of thousands capacity which the vendors claim. I find them a little exaggerating. The OLT also has the ability to isolate VLANs with transmission containers(T-Conts)and places traffic prioritization on them. That's one way how they do garden walled IPTV services like FIOD TV.
No doubt many call the OLT a dumb Layer 1 switch that has most of the upper layers taken care of by the core routers such as IP routing, but there are some tasks which is unique to the OLT.
The real work on FIOS is being done by the routers sitting behind the OLT boxes.
This is the 2nd weakness of the GPON OLTs I pointed out earlier. The model itself is best suited to serve high density populated areas with many lines per port (>30 splits per port) to maintain its competitiveness against active switches. Now here's the catch: You buy a very expensive OLT set that claims to cater for 10s of thousands of ONTs after receiving all the demand from the locals. But your backplane of your OLT only takes in multiple 1GE, 10GE and 100GE at most.
How do you intend to keep up with the growing demand?
The key point is to learn from the cable industry where they too had faced this very same problem with the headends not keeping up with the rate cable modems installations. Many started complaining of slowdowns during peak hours and they are not getting close to their subscribed speeds.
Eventually they learn from their mistakes and solved the issue by creating more "mini nodes" to spread things out more evenly by means of effective rationalizing.
In the case of passive networks the same move can be done by increasing more OLT ports, buying more miniaturized OLTs and decreasing split ratios. But before you do that, don't you think the architecture of active switches were better suited for such situations? You are better organized and ready to serve with packages beyond 1gbps with dedicated active ports than you do with passive networks.Would you buy an expensive cabinet sized OLTs just to do few hundred lines? Better spend your extra on the core routing and transport.
I can count my dedicated active ports better with a 10GE than an expensive OLT with a single GPON line card with many logical connections over shared lines.
A monster still needs a matching digestive tract to suit its appetite.
We have to consider our options resonably that 1GE and 10GE are the only mainstream options for uplinks. 100GE is ready but equipments are still expensive. Why would you spent on a big box that claims a capacity which its backhaul can't keep up?
Fiber is not an issue with economics of scale. It's going to get cheaper since silica is plenty to be found in the ground. Manufacturing process improvements will only lower manufacturing cost over time till it becomes cheaper than copper one day. For new ISPs building new infrastructures, it's best they keep solely to Ethernet to simplify their setup. TDM and ATM should be phased out entirely for modern NGNs.With IPTV, very few service providers are using the RF overlay option for their cable tv services.