|
This overview will discuss core network technology and cost trade-offs inherent in choosing between "analog" architectures with high optical transparency, and ones heavily dependent on frequent "digital" signal regeneration. The exact balance will be related to the specific technology choices in each area outlined above, as well as the network needs such as node geographic spread, physical connectivity patterns, and demand loading.
Over the course of a decade, optical networks have evolved from simple single-channel SONET regenerator-based links to multi-span multi-channel optically amplified ultra-long haul systems, fueled by high demand for bandwidth at reduced cost. In general, the cost of a well-designed high capacity system is dominated by the number of optical to electrical (OE) and electrical to optical (EO) conversions required. As the reach and channel capacity of the transport systems continued to increase, it became necessary to improve the granularity of the demand connections by introducing (optical add/drop multiplexers) OADMs. Thus, if a node requires only small demand connectivity, most of the optical channels are expressed through without regeneration (OEO). The network costs are correspondingly reduced, partially balanced by the increased cost of the OADM nodes. Lately, the industry has been aggressively pursuing a natural extension of this philosophy towards all-optical "analog" core networks, with each demand touching electrical digital circuitry only at the in/egress nodes. This is expected to produce a substantial elimination of OEO costs, increase in network capacity, and a notionally simpler operation and service turn-up.
At the same time, such optical "analog" network requires a large amount of complicated hardware and software for monitoring and manipulating high bit rate optical signals. New and more complex modulation formats that provide resiliency to both optical noise and nonlinear propagation effects are important for extended unregenerated reach. More sophisticated optical amplifiers provide lower noise for increased reach and increased spectral bandwidth for higher wavelength count lower wavelength blocking probability. Optical analog networks also require methods for mitigating optical power transients, for controlling optical spectral flatness, and for dynamically managing changes (e.g. in chromatic dispersion and polarization mode dispersion.) Since signals stay in the optical domain, optical performance monitoring techniques are required for fault isolation and correction. Efficient routing of optical signals also requires sophisticated switching nodes with an ability to selectively steer optical signals into several directions with single-channel spectral granularity. Most of these technologies are not modular and require interruption of service if not deployed at the initial system installation, thereby increasing first install costs substantially, even if initial capacity loading is small.
Further, validation of systems and software targeting a specific network design is complex. Only a small fraction of the total network may be reasonably reproduced in the lab, and many field configurations are not predictable or even dynamic. Thus, extra system margin has to be allocated to handle the behavior uncertainty.
To constrain the complexity of both hardware technology and software algorithms, regions of network transparency can be established with OEO forced at perimeters. Thus, "analog" regions are surrounded by "digital" interfaces.
Following are some example tradeoffs that will be discussed. What is a good modulation format choice, and does increased reach and impairment resiliency justify hardware and controls that are more complicated? What are reasonable amplifier choices to make under specific network assumptions? Can high transport system capacity be leveraged to simplify optical switch node design by reducing spectral efficiency?
|
