Maybe you can get Morgan Stanley's team of telcom experts to explain why running a dslreports.com speed test to various servers across the country yields latency as a function of distance. Hops, baby, hops, which is a switch issue, or is it? Maybe it is better to uplink the spectrum to satellite and then downlink given all these hops. If so, I'd say we have more of a 'bone issue than was previously assumed, because all the switches in the world can't make all those hops and their processing millisecond delays go away.
They say,"we have lots of dark fiber", so there is no 'bone issue. It's dark because the assumption of the marginal cost to light it isn't justified since the marginal cost of traffic lost due to hop switch latency is higher. Why light it if the intermediaries render its employment impractical?
Or they say, "we have the SONET ring structure in place which can be enhanced". The question is whether SONET can handle the growth in BB and at a reasonable cost. It can't.
Until now the structure could handle the circuit switched setup, but cable BB is suddenly and quietly taking off. People's prejudices against it which was well-orchestrated by circuit switch legacy equipment preferring non-competitive RBOCs is now fast disappearing. The people are dumping DSL for cable. With BB ramping you have an exponential demand for bandwidth whose supply requirement must be distributed across the entire nation.
It boils down to this: they had better start re-doing the major trunk model because metro ain't going anywhere with the hop situation as it is. Traffic isn't mostly local.
At least a hierarchical model which employs unencumbered long distance 'bones is needed. Another modelling possibility is to employ an optic mesh model as conceived by SCMR. PONS with purely optic, non-MEMS, switches in this model enable it to avoid the long distance hop latency, and so dedicated long distance wires aren't necessary. |
