Ken, Jack,
The issue concerning 1- or 2- strand configurations is interesting, and deserves a bit more discussion, IMO.
After all aspects of service provisioning are taken into account, including restoration modes, single-strand configurations are probably better suited to the long haul. This is due, in part, to the economics associated with IRU/leasing and the overall higher costs involved as a measure of mileage and the requirement for optical amplification (and 3R regeneration) at regular intervals.
First, there are tradeoffs, and some of them are changing, dynamically, as we type, making a single strand alternative attractive, but not quite yet for most applications - especially those applications where packing the highest number of lambdas "on a full-time basis" is critical.
Secondly, when a single strand is used, as in some dual-strand alternatives, a second strand is usually leased from a second provider, or on a diverse route, for fail-over purposes, anyway. That is, if the provider intends to guarantee service availability.
In the end, and depending on the topology used, there is no free lunch. When all is said and done, a 2 fiber system still supports twice as much bandwidth "on a full time basis" as a single strand system does, especially when ring topologies are used. This holds true especially in ring topologies, but it is not necessarily true for the mesh, since mesh topologies can (or will, in the future), in theory, provide improved efficiencies across all modes of operation, compared to ring modes.
But mush isn't quite there, just yet, at least not on an ubiquitous or mutli-provider level, by any means. So far it works best, where it is used at all, on isolated networks that are unique to a single provider, or between two providers with internetworking agreements, most often when they are using the same vendors' platforms.
So, the question centers on the capacity requirement and the topology in question, and the means used by that topology to recover itself in the event of a fiber or node failure.
In short-haul ring networks,** the fiber costs are considerably lower as a percentage of overall system costs than in the long haul.
** Yes, even most next-gen metro networks gravitate to a ring topology, of one form or another, for reasons having to do with a highly-simlified (if not elegant and sometimes costly (due to preemption), from a bandwidth perspective) self-healing capabilities, and the ease by which they avail themselves to subscriber insertion along existing routes.
Two strands usually cost less as a percentage of total cost of ownership in the metro; they do not require amplification at regular intervals as do their longer-haul brethren; and two strands provide a more economical means of 'wrap-arounds' for ring restoration purposes during failures than do single strand configurations that must depend on another level, or two, of lambda executions to achieve the same end. Let's say that the lambda option of restoration were equally efficient, though.
Think of the consequence on a 1-fiber system. The following is an over-simplification, for illustrative purposes, only, although something very similar takes place:
In a 1-fiber system, if two bands were used in each direction for both transmit and receive in the "working mode," then, in the restoration mode, one band would continue to be used in the transmit direction, but the second would now become the return path for the first.
In this case all users who had been riding over the second band during the "working mode" (i.e., during normal operation) would have been preempted, or bumped, during the restoration mode, because the band they were working on was needed to make good those who were riding over the guaranteed service on the first band.
To avoid this ugliness, many SPs (including some in the SONET regime) do not lease those secondary pre-emptable channels, and when they do those channels are available on a contingency, sometimes- discounted basis, only, with the understanding that they might get bumped.
----
Exp: In the event of a loss of signal, most wdm ring networks go into a loop back mode, in order to bypass the failed section. Effectively, this creates an elongated ring, instead of a circular one.
This is done in a way that is similar to SONET and FDDI rings. It is still far less costly [at the network element level] to effect this kind of loopback at the physical strand layer (all that's required is a sensing circuit and a physical latching mechanism), when you have two strands to work with, than to perform a similar loop back at the optical (lambda) levels.
Although the optical latching mode is do-able, too, but even when the latter optical mode is used, as with the physical 2-strand mode, the same kinds of tradeoffs rear their heads up as those that appear in certain SONET ring configurations. These have to do with half of the available lambdas being relegated to "standby" or "preemptable" status, instead of being available on a dedicated, full-period basis.
Note: In SONET circles (pun intended), a two fiber ring, as opposed to a four fiber ring, fosters the same kind of evaluation and trade-off decision making as a 1-fiber wdm and a 2 fiber wdm configuration, respectively. The two-fiber SONET ring option requires that half of the OC-192 (i.e., fully 48 T3s out of a possible 192 T3s) be held in standby or throw-away status (which means that users on those standby T3s get "bumped" if there is a failure), in the event of a fiber break. On the four wire SONET option, on the other hand, all 192 T3s continue to function normally, with the exception of the 50 msec hit that occurs during the initial break. And some systems even mask the latter 50 msec hit through buffering techniques.
It's a matter of cost-benefit concerning state-of-the-art optical element costs versus mileage topology considerations, as always. But like I said, there is seldom a free lunch. As the intricacies of mesh designs take hold over time the calculus will change. But there are mountains to move, before then.
I've actually over-simplified this, considerably, because a four-fiber system can provide even more "full time" (non-preemptable) bandwidth than one with only two fibers. The same principles apply. But at some point the question has to be asked: What price maximum capacity-availability? Comments and corrections are always welcome.
FAC |
