With the exception of one assertion that you made concerning the survivability of rings (the last point that I address below), each of your points is useful because they stimulate thought and challenge some of the sacred beliefs that are based on past and existing, but not necessarily future, realities. In some ways you've captured the essence of some capabilities that might some day hold true, but, for the moment, are misplaced in pruely temporal way, because they are not yet relevant in the realm of real world subscriber delivery schemes today.
As an example, the eventual densities of certain forms of mesh delivery to user populations. Again, these might become so, over time. I say 'might become so' because many of the emerging mesh, and even ring-based technologies, such as resilient packet rings <although rpr is well on its way to acceptance, I believe, despite some questions that are being asked about it> that we are discussing here are still nascent in many ways, and their future with respect to "which" venues and which applications that they will eventually serve - if not the extended viability of the plaers who support them, in general, witness the difficulty that the BLECs have had - remains to be seen.
For the emerging technologies supporting mesh topologies that succeed, they may spread like the blob, and they may become as ubiquitous and pervasive as the delivery of electric power. In those instances your views concerning the ease of mesh restoration might become realized. However, even if packets supported by muplitple service providers flow in a vast maze of meshed links like electricity throughout a MAN, or even throughout a neighborhood or building, we still have to be mindful of the logical and physical domains of connectivity of each provider, within which each service provider will limit both its own reach and its restoration capabilities. In many of these situations their delivery will be single threaded, which accounts for why they can compete: single threaded delivery over a single strand is a lot cheaper to provide than delivering SONET over four strands.
To clarify, SONET rings typcially employ 4 strands to a location <two in and two out>, although other configurations can call for as many as 8 <four in and fout out> or 2 <one in and oue out, depending on the type of ring that it is). In contrast, an EPON might serve an entire building or campus through the use of a single unprotected strand, through the use of passive star couplers. Yes, you can protect by adding redundancy and multiple paths of egrees/ingress, but the more one adds redundancy the more one removes one of the primary incentives (read: the price attractiveness) that would cause a prospect to be drawn to it [single stranded links hanging off of passive star couplers] in the first place. But this is doable, and it may begin to catch on with greater regularity if the technology take up is great enough, and if the fiber resellers and integrators (including BLECs) live on to make it so.
In any event, your points warrant further discussion, IMO. In the meantime, the folloing new tutorial on the IEC Consortium Board might be of interest. It's from Alloptical, one of the supporters of the EPON space and a member of the EFM Study Group. It gives some good historical perspectives with respect to the evolution of EPON, as well as providing some good insights into the technology, itself.
Note in the first panel where the tute states that EPON has borrowed from the ATM over PON (APON) initiatives that are a part of the original FSAN model, which is another topic we recently discussed here, several days ago.
iec.org
To address your statement concerning the survivability of rings in the event that a fiber cut were to occur on any of the ring's sections:
"Rings don't provide indefinite reliability because if the ring trunk is cut, there is no alternative path."
... see Fig. 8 in the following paper from Cisco:
cisco.com |
