Economic analysis of high-capacity architectures in metro applications
Economic analysis of high-capacity architectures in metro applications
Contrary to conventional wisdom, OC-192 4-fiber BLSR can save 15% over OC-48 metro DWDM in equipment costs, time to market, engineering and installation costs, and footprint.
Steve Carter, Hitachi Telecom (USA) Inc.
Since OC-192 transmission speeds began making inroads in long-haul networks in 1996, 10-Gbit/sec technology has become widely accepted, with most new carriers implementing it from the beginning. And though dense wavelength-division multiplexing (DWDM) was originally touted as a technology that would make OC-192 unnecessary, now OC-192 installations routinely include the provision for the concurrent or future implementation of DWDM, opening the possibility for transmission at 320 Gbits/sec and beyond. Network complexity is significantly reduced at these high bandwidths as compared to OC-48 (2.5 Gbits/sec). For example, at 320 Gbits/sec, 32 OC-192 network elements--or network layers--are required, while the same 320 Gbits/sec at OC-48 would require 128 network layers.
Although OC-192 has some disadvantages, such as a lower tolerance for chromatic dispersion and polarization-mode dispersion, compensation techniques as well as new fiber types have been developed to offset these problems. And now that 4-fiber bidirectional line-switched ring (BLSR) capability is becoming available in OC-192 equipment, the combined advantages of Synchronous Optical Network (SONET) protection, higher time-division multiplexing (TDM) rates, and DWDM are becoming very attractive for metropolitan backbone applications.
To support this assertion, an economic comparison of OC-192 and OC-48 in a metro network will follow. It will not only show that OC-192 offers as much as a 15% cost savings but that the OC-192 solution enables much greater usage of the gross available bandwidth as well.
Tale of two networks
This analysis looks at a typical metropolitan application: a generic 5-node ring of less than 50-mi circumference with normal traffic growth. Network expansion is modeled through 12 stages of typical traffic growth. The following two scenarios are compared:
OC-48 2-fiber BLSR with metro-optimized DWDM. Expansion is achieved by adding layers of OC-48 capacity through DWDM.
OC-192 4-fiber BLSR with subtending OC-48. Expansion is achieved by adding circuit packs to provide the tributary capacity required.
The following factors are critical to this evaluation:
Equipment cost (the cost per bit of OC-192 is approximately 38% less than the cost for equivalent capacity based on OC-48)
Time to market (unserved demand equals lost revenue; time required to implement upgrades)
Footprint (space, power, and maintenance)
Engineering and installation costs
This economic analysis uses both objective assumptions such as market pricing for equipment and revenue, and several practical assumptions that must be considered to achieve a "real world" result (see Table 1). One such practical consideration, "time to market," is an often overlooked cost of upgrading a network. The time required to order, receive, engineer, and install new equipment can be significant, and during that time, the new equipment cannot generate revenues.
An example: A service provider receives an unforecasted order for OC-12 service. If an OC-12 circuit is not available immediately, revenue is lost. For every day that it takes to turn up the circuit, revenue losses continue to accrue. Excessive delay may cause the customer to seek another service provider, resulting in permanent loss of the potential revenue.
The following additional assumptions are included:
The price of tributary cards is assumed to be equal and not included in the study.
Likewise, the price of spares is assumed to be proportional and is not included.
Training and test equipment requirements are not included.
Since growth rates vary tremendously, net present value computations are not included.
Survivability of 50 msec is assumed for both cases, as is reasonable market pricing.
Uneven demand growth
In practice, growth most often occurs unevenly, between two or three sites, which presents a problem. In a ring architecture, survivability depends on the ability to reroute traffic in either direction around the ring. Thus, the line capacity of the entire ring must be upgraded to accommodate the maximum traffic characteristics between individual segments. However, as Asynchronous Transfer Mode (ATM) and Internet Protocol (IP) applications continue to multiply, so does the demand for greater network reliability for "mission critical" traffic. Thus, the appeal of SONET ring architecture for its survivability capabilities increases. For this study, we assumed a 12-stage pattern of growth (see Fig. 1).
As we work through the demand growth stages, we begin to encounter the limitations of the OC-48-based network. For example, in implementing stage 5, we cannot route the OC-48c (2.4-Gbit/sec) traffic because the maximum line capacity of a layer (wavelength) on the 2-fiber network is only 24 DS-3s (44.736 Mbits/sec). The primary disadvantage of the OC-48 solution is that the line capacity of 24 DS-3s limits the usage of tributary slots (see Fig. 2). In some cases, only half of the tributary slots are available because of this limitation. Finally, since OC-48c cannot be carried on a 2-fiber OC-48 plant, service providers must limit their customers to OC-12c (599.04 Mbits/sec) or below.
In the OC-192 4-fiber BLSR configuration, only one ring is required, supplying a line capacity of 192 DS-3s. The time-slot assignment capability of the OC-192 equipment allows the subtending OC-48 multiplexer shelves to fill optimally. An additional advantage is that the OC-192 ring model could be as large as 400 km in circumference because the optical signal terminates at each node. Low-cost metropolitan WDM solutions are limited to between 80 and 100 km for the entire ring circumference, while the OC-192 solution allows 80 to 120 km for each link.
Line and tributary capacity characteristics
Although capacity appears to be adequate when we look at the gross capacity of the system, no single OC-48 system can meet the demand (see Fig. 3). Additional wavelengths--and equipment-- must be provisioned. The case of four OC-48 systems and a single OC-192 can be compared to a four-lane highway. In the first case, the highway has barriers between each lane, preventing vehicles from changing lanes. In the case analogous to OC-192, vehicles can change lanes as needed, so that the capacity of all lanes is available to all traffic.
One important reason for performing this "real world" analysis is that many situations are observed that are not intuitively obvious to predict. For example, it turns out that the OC-48 network is an order of magnitude more difficult to engineer. In many cases, traffic demands must be split and routed across multiple OC-48 rings to efficiently meet demand. In other cases, the demand simply cannot be met, although capacity appears at first glance to be adequate (see Fig. 3).
Two significant examples of such situations warrant a detailed observation: DS-3 efficiency and OC-48c transport capability.
Planners familiar with OC-48 2-fiber BLSR networks will readily agree that the average node on the ring rarely reaches a full 48 DS-3s. In fact, on average, each node can be considered at maximum capacity with 24 DS-3s or less. This limitation results from the fact that the line capacity on a 2-fiber BLSR system always exhausts before tributary capacity. For example, in the case of a 5-node ring, the total tributary capacity on the ring is 48 ¥ 5 = 240 DS-3s. At peak efficiency, the line capacity is 120 DS-3s (note that each DS-3 has a termination point at each end, so 120 DS-3 lines = 240 DS-3 tributaries). However, the real available line capacity is a function of the actual traffic demand pattern, and in practice, an OC-48 2-fiber BLSR ring rarely achieves more than 60 DS-3s. Thus, although the service provider paid for equipment capable of terminating 240 DS-3s, only about 120 of them can actually be used. The undesirable revenue/cost ratio is obvious.
With an OC-192 4-fiber BLSR, DS-3s must be input from a subtending OC-48 multiplexer (see Fig. 4). At first glance, this arrangement appears to be inefficient. However, it allows grooming at the OC-192 level and is much more efficient, because a subtending OC-48 multiplexer can support a full 48 DS-3s. The OC-192 system provides the time-slot assignment function and the high-speed bandwidth conservation for ring switching. The end result is the OC-192 network requires about half the number of OC-48 terminals to terminate a given number of DS-3s. Figure 4 shows this scenario at a typical node. The OC-48 network requires four add/drop multiplexers to terminate 96 DS-3s, whereas the OC-192 network requires only two.
As far as transport of OC-48c signals is concerned, the debate over the most efficient method of transporting packet-switched data has revealed two truths:
1. Packet-switched data will soon become the dominant bandwidth consumer in the network.
2. Competition is forcing carriers to find cost-effective, yet resilient ways to transport data traffic.
In response to this overwhelming market need, IP and ATM switch vendors are now offering SONET-framed OC-48c interfaces directly off their equipment. This capability enables 2.4 Gbits/sec of data traffic to be put onto the SONET backbone through a single interface. Simply put, this is not possible unless the backbone is OC-192. An OC-48 2-fiber BLSR backbone can transport a maximum of OC-12c. With this limitation, an OC-48c must instead be transported as four OC-12cs, resulting in a 40% cost penalty at the optical interface.
One alternative is to port the OC-48c directly to the WDM layer, completely bypassing the SONET multiplexer. But carriers are quickly realizing that many benefits such as bit-error-rate performance monitoring and 50-msec switching are requirements only available via the SONET multiplexer. As a result, it is becoming commonly accepted that OC-48c cannot be transported on an OC-48 network.
Table 2 details the OC-48 DWDM and OC-192 4-fiber BLSR solutions in each category. Individual percentages contribute to the total 15% cost savings.
The OC-192 4-fiber BLSR is shown to achieve 15% overall savings over OC-48 metropolitan DWDM in the outlined application. One OC-192 4-fiber BLSR system serves the equivalent of 11 wavelengths of OC-48. The STS-1 (52-Mbit/sec) granularity of the OC-192 time-slot assignment matrix provides optimal efficiency of subtending DS-3 shelves. Equipment costs, time to market/unserved demand costs, engineering/installation costs, and footprint all contribute to the overall 15% savings.
In addition to cost savings, the OC-192 solution provides an optimum blend of network simplicity and capacity for metropolitan applications. DWDM is not eliminated from the equation; rather, it is simply brought in at a higher TDM rate, allowing network operators to achieve even higher bandwidths in the future, while minimizing increases in complexity. u
Steve Carter is senior manager, account marketing, at Hitachi Telecom (USA) Inc. (Norcross, GA).