Terabit super-channels and multi-layer switching architectures enable optimal capacity allocation and network efficiency to meet evolving network dynamics. Anuj Malik, senior manager of product and solutions marketing, and Gaylord Hart, director for CATV market segment of Infinera, explain how.
Around 150 million kilometers of new optical fiber are produced each year, and there is many times that already installed since 1997. Network operators must ensure maximal return on their investment, as well as future proofing new installations in the face of colossal and rising bandwidth demand.
Significant progress has already been made in core networks by migrating from 10 Gb/s to 100 Gb/s optical transport waves that allow 8 Tb/s or more of capacity to be carried on traditional fiber using the standard 50 GHz C-Band ITU-T G.694.1 grid. However, bandwidth growth projections already suggest that soon even 8 Tb/s per fiber will no longer be enough. What is more, the operational costs of deploying so much capacity in 100G increments can be high. Network providers must be prepared for ever-higher capacity demands, and they must increase operational efficiency to meet these demands in a flexible, timely and cost-effective manner.
This article outlines why next generation optical networks are needed to replace today’s rigid channel structure with a flexible grid of variable-width optical super-channels. These terabit scale super-channels can then be implemented to meet appropriate modulation and reach requirements, with provision for software controlled optical switching.
EFFICIENT USE OF BANDWIDTH
To make best use of fiber spectral capacity and lower CapEx, most metro and long-haul optical networks use Dense Wave Division Multiplexing (DWDM) to transport multiple waves over a single fiber. The ITU standardized a fixed DWDM channel plan with built-in guard bands between each optical channel to allow for multiplexing and de-multiplexing of individual waves for routing as well as filtering waves at their destination. These guard bands occupy up to 25% of spectrum, which amounts to a loss of capacity.
The industry is now migrating to optical super-channels, which are much wider than the traditional ITU-T grid channels, but which have no internal guard bands between waves.
Figure 1 shows the difference: on the left there are 12x100G waves using the standard grid with guard bands highlighted in red. So 1.2 Tb/s of transport capacity requires 600 GHz of optical spectrum for transport. On the right is an equivalent multi-carrier super-channel also with 12x100G waves. Because the super-channel is switched or multiplexed/demultiplexed as a whole, no internal guard bands are needed, only those at the lower and upper edges of the super-channel shown in red. So the same 1.2 Tb/s capacity only needs 462.5 GHz – a 23% reduction of optical spectrum (comparable to the saving when you buy one economy-size pack instead of twelve mini-packs). From an optical channel perspective, it is the difference between twelve discrete 50 GHz channels and one 462.5 GHz super-channel.
This is just an example, as super-channels may be implemented in many other ways to support up to 24 Tb/s per fiber – Figure 2 shows some alternatives.
The single-wave super-channel on the left is the simplest to implement with the fewest number of components, but to support 384 Gbaud it requires ultra-fast silicon that might not be available for 8 years or so. A single-wave super-channel also allows no flexibility in allocating or routing bandwidth with smaller granularity since it consists of a single indivisible wave.
The dual-wave super-channel in the center is only a bit more complex, with just twice the components, but still needs 192 Gbaud electronics that might not be available for another 5 years. It is also restrictive, but it does allow the two waves to be configured and handled with a choice of one integrated channel or two separate channels.
The 12-wave super-channel consisting of twelve 100G waves shown at the right does require twelve times the components of the single-wave version, but it does operate at speeds supported by today’s silicon technology. It also provides greatest flexibility, since individual waves may be combined in any permutation, and modulation formats can be assigned on a wave-by-wave basis to further reduce CapEx.
Most network operators would put up with this greater complexity now and enjoy higher capacity, rather than wait several years in the hope of a simpler solution. Photonic integration, however, makes it possible to implement all the necessary components for a terabit scale multi-carrier super-channel on a pair of Photonic Integrated Circuits (PICs), one each for transmit and receive functions.
Third generation PICs supporting 5x100G super-channels aligned with the ITU-T fixed channel grid have already been deployed in the field for over two years. They integrate over 600 optical functions on a pair of chips, to replace over 100 discrete optical components and over 250 fiber couplings – with significant improvements in density, power consumption, heat generation, and reliability. So PICs can bring the component count back in line with the single-wave implementation as the most practical path to scaling optical networks in the future.
BALANCING REACH VS. SPECTRAL EFFICIENCY
Higher order modulation formats make more efficient use of spectrum, but are more susceptible to noise and cannot reliably reach as far. For example, 16QAM with 4 bits encoded per symbol is spectrally twice as efficient as QPSK with 2 bits encoded per symbol, but its reach is about a quarter of that of QPSK.
Provisionable modulation would allow each channel to be optimised for reach versus spectral efficiency, and so deliver greater cost savings, but it would require a flexible grid system to support variable bandwidth channels. The ITU-T’s latest WDM grid specification, G.694.1, has defined a flexible grid with WDM channels having a 12.5 GHz width granularity in place of the coarser 50 GHz width in the fixed grid. This flexible grid lets the provider define an aggregate super-channel in multiples of 12.5 GHz to accommodate any combination of optical carriers, modulation formats and data rates, in order to balance spectral efficiency against extended reach of the optical signals.
In addition, the flexible grid makes it possible to allocate frequency slots and modify modulation formats to meet changing to traffic demands. This allows resources to be used efficiently in response to traffic variations. Over the past two decades the bit rates and modulation formats of optical transmission systems have evolved dramatically, but it is increasingly difficult to raise the bit rate while maintaining spectral efficiency and reach. This flexible grid system will allow operators to install line systems today that will accommodate virtually any type of super-channel tomorrow– and so future proof today’s capital expenditure.
Network planners need to consider not just capacity, but also the mix of service types. Even though the line side transmission rate is evolving beyond 100 Gb/s, more than 95% of the client services in the network are still 10G or less, with muxponders being used to aggregate these services onto 100G wavelengths. But 100G muxponders do not provide any ability to groom or switch sub-wavelength traffic within or between wavelengths. This can lead to low utilization of deployed bandwidth and therefore over-deployment of 100G wavelengths, also termed the “muxponder tax”. The digital switching architecture solves this problem by efficiently grooming client services into line side but as the demands become fully-fill, it makes more sense to use optical switching for low cost. Next generation optical network should support a multi-layer switching architecture that integrates digital and reconfigurable optical switching. This combines the benefits of a digital switching for sub-wavelength grooming and optical switching for operational simplicity and flexibility for express traffic.
Figure 3 illustrates the difference. In optical-only switching, muxponders aggregate low speed traffic onto the higher speed line side waves switched as a whole – with no ability to add/drop traffic at intermediate nodes. This is very efficient, given sufficient traffic to fill the line side wave. With digital-only switching, low speed traffic is efficiently aggregated onto line side waves, but all the service traffic is routed through intermediate nodes, even if it is not being added or dropped at those nodes. Digital switching is efficient for filling line side waves and add/dropping traffic from those waves at multiple add/drop sites.
Using multi-layer switching, digital switching can efficiently groom traffic onto line side waves to reduce stranded bandwidth, and these waves can then be directly routed optically to those nodes where client traffic is added or dropped. This uses the minimum number of wavelengths under both extreme conditions: either when demand is low and the wavelengths are not filled, or when demand is high and the wavelengths can be highly filled.
THE WAY AHEAD
The unceasing demand for additional optical transport bandwidth is driving new technology at a very rapid pace to increase spectral efficiency and bandwidth utilization, all while lowering the overall transport cost per bit. Multi-carrier super-channels (which operate on a flexible grid channel plan that supports variable bandwidth channels) increase spectral capacity by eliminating the inefficient guard bands associated with fixed grid channel plans and by enabling provisionable modulation formats, which allow operators to configure their networks for optimum spectral capacity versus reach. 16QAM modulation can deliver up to 24 Tb/s of capacity per fiber, but at the expense of much shorter reach compared to QPSK.
To support multi-carrier super-channels, a flexible grid line system is required which allows channel and switching bandwidth to be assigned as needed to each individual variable bandwidth super-channel. The flexible grid architecture supports bandwidth assignments in the C-band in 12.5 GHz increments, allowing efficient use of the C-band for both fixed and flexible grid channels.
To achieve maximum transport efficiency with super-channels, which have much higher bandwidth than 100G fixed grid waves, a multi-layer switching architecture with integrated optical and digital layer switching is required. This architecture optimizes super-channel bandwidth utilization by allowing sub-lambda OTN digital grooming within and between super-channels, and optimizes super-channel routing between destinations with flexible grid optical switching. The resulting “optical data plane” is also ideally suited to integrated Control Plane operation, whether that is GMPLS today, or a Carrier SDN Control Plane in the future.
- Anuj Malik & Gaylord Hart, “The Evolution of Next-Gen Optical Networks: Terabit Super-Channels and Flexible Grid ROADM Architectures”, SCTE Cable-Tec Expo 2014 Proceedings
- Soumya Roy et.al, “Evaluating Efficiency Of Multi-Layer Switching in Future Optical Transport Networks”, OFC 2013