Metro, regional, long-haul, metro-access, metro-aggregation, metro-core, ultra-long-haul, data centre interconnect… whatever these terms mean to you, I can almost guarantee that we would disagree somewhere in our views of exactly what these terms mean and where specifically these products are used in optical networks. Our expectations of exactly what distances these systems would cover and the functionality that each should have would probably also vary considerably.
There used to be a clear difference between metro and long-haul WDM systems without the complication of sub-groups. But this hasn’t been the case for a few years and the lines are becoming increasingly blurred as optical transport systems continue to evolve. Metro systems have expanded to encompass a non-homogenous set of definitions, with high-end metro-core systems encroaching on the space traditionally held sacred by long-haul platforms, while metro-access systems push deeper and deeper into access networks.
Although it might be tempting to see this as an opportunity for a ‘one-size-fits-all’ product strategy covering both metro and long haul, the technology, performance and economic demands within today’s networks are demanding dedicated products for both markets and, in some cases, even application-specific products.
The introduction of metro
Back in 2004, when I started working for Transmode, the company was still a metro coarse-WDM (CWDM) specialist start-up. The segmentation between metro and long-haul products was quite straightforward at that time. The bulk of WDM hardware shipped was for long-haul systems, which addressed transmission over regional, national and international networks.
The metro WDM market was adequately supported by simple CWDM or DWDM systems, which started out as lower-cost, simplified versions of long-haul WDM transport systems. Some were created by stripping down long-haul systems to reduce expenditure while others began with a clean sheet of paper and built new metro-focused systems without the constraints of the long-haul system architecture they had been previously using. Those systems have evolved, adding new functionality and improved optical performance over the intervening years.
There have been winners and losers in this market and, in my view, those that are winning today are companies that started with that clean slate and a real metro perspective of the basic product architecture. Systems vendors that took a real metro vision from the outset have really been able to focus on areas such as low power, high density and simplified operation/ease-of-use. Being brave enough to limit power in chassis to the levels needed for metro only, and to make tough decisions around what functionality you can dispense with in the metro have been key.
What is metro anyway?
A large part of the confusion over what constitutes ‘metro’ is that there are two major factors that determine whether a system is considered metro or long haul. The obvious one is the distance supported in the optical path, while the sometimes less-obvious factor is functionality, whether it suits network edge or core.
A definition according to transmission reach ought to be quite simple; however, each person will have their own perspective of metro reach vs long haul reach, that could mean anywhere from a few hundred kilometres to more than 1000km for regional networks deployed using metro systems, especially as we move to coherent optics for 100G and beyond. Even more unclear, however, is the categorisation by functionality, and I believe this leads to the biggest confusion over the various terms and categories within optical networking, as I will explain.
Today’s metro WDM systems vendors have used component technology advances to evolve their platforms to support functionality such as greater distances, 100G wavelengths, ROADM-based optical switching, OTN transport and aggregation/switching, Ethernet/MPLS-TP switching and service provision, often within the same simple chassis that were deployed back in the early days of metro WDM.
So if metro systems are getting closer and closer to long-haul systems, why doesn’t a one-size-fits-all approach work? To answer this question, we firstly need to compare the major tasks of metro and long-haul networks. Metro networks obviously deal with shorter distances; however, this can be anything from metro access with direct fibre access to the end customer through metro aggregation to metro core. Probably the biggest differences between metro and long-distance networks are service awareness and bandwidth granularity.
Metro networks carry many different services or traffic types, often over the same single platform and physical infrastructure. Many of these services require important transport characteristics to provide basic operation of the service or differentiation of the service against competitive service providers. This can include characteristics such as:
- Service type – Ethernet, SDH/SONET/OTN, Fibre Channel, CPRI, etc;
- MEF compliance for Ethernet services;
- Multi-cast, E-Tree or other non-point-to-point service type for video or enterprise services;
- Low latency for financial services or mobile fronthaul; and
- Synchronisation for mobile backhaul and fronthaul services.
Contrast this approach to the core of the network where the primary purpose is highly economic point-to-point bulk data transport with a high degree of flexibility and scalability but much lower, if any, service awareness requirements. These systems and networks will deal with high levels of bandwidth and will focus on the transport aspects without necessarily needing to consider the specifics of individual services.
The need to be service aware has a huge impact on the success of a metro WDM or packet-optical platform with those platforms that are doing well typically having strong capabilities in this area. The better these platforms perform, the better chances the network operator has to differentiate their products and services within their marketplace.
The network lives forever
As metro WDM networks expand and spread more deeply into access networks and into metro-core segments of the network, then the number of nodes and the physical size of the network become significant factors. Long-haul networks typically deal with a smaller number of higher capacity nodes and this creates more of an opportunity to replace long-haul networks when a step change in technology occurs.
This happened most recently in long-distance networks with the introduction of new 100G technology over the past few years. The new technology made it economical in many cases to totally replace existing 10G networks with new 100G networks in what Andrew Schmitt of analyst firm Infonetics Research (now part of IHS) called the 100G Optical Reboot. This network replacement may also happen in some parts of the metro network, but unless there is something fundamentally wrong and limiting in a metro network then the operator will want to add new technology to the existing network rather than swap out the network with a brand new system.
When 100G was introduced in long-haul networks, new systems and networks dedicated to 100G transmission were typically created and the older 10G systems were retired. A similar approach is often impossible or undesirable in the metro due to the larger installed base. The majority of the metro network will need to migrate gradually from 10G to 100G transport, while maintaining legacy services and adding new 100G bandwidth to high-demand routes or new 100G services directly to customers on a case-by-case basis.
This trend leads to a number of differences in how 100G is handled in the two networks. Using 100G in the long-haul sector allows carriers to take advantage of coherent detection’s better tolerance of non-linear effects and remove the need for dispersion compensation from the network, for example, or benefit from new gridless technology, using it to create super-channels for better network capacity utilisation.
Meanwhile in the metro, 10G transport will play a significant role in new bandwidth deployments for many years to come. It constitutes the bulk of the historically-deployed bandwidth that itself will only gradually migrate to 100G transport. As a result, metro 100G systems need to be designed in a way that supports this hybrid environment of mixed 10G and 100G systems, and the move to other emerging technologies such as gridless filter architectures. Such development needs to consider both new-build gridless environments within the network and legacy hybrid environments, where the cost of replacing the whole optical infrastructure for the few initial 100G links just isn’t viable.
Dedicated platforms for DCI
One area within the metro where new, dedicated systems will be viable is for data centre interconnect (DCI). Some vendors have now introduced new highly optimised 100G or even 200G-based dedicated DCI platforms that address this application, providing extremely high-capacity point-to-point connections. These new systems, which are built to the very specific requirements of the large internet content providers and mega data centre operators, have the potential to address a good segment of the DCI market.
However, there are many other DCI applications that can be supported by lower rates or a mix of lower rates with 100G transmission over metro-type infrastructure: for these functions, protocols such as Fibre Channel and 10G are still important, and network topologies other than point-to-point become applicable, such as ring-based networks. These applications include traditional telecom service providers and enterprises such as banks and universities.
While there is a case for application-specific metro systems for DCI, we shouldn’t assume that all future DCI projects can be addressed by these systems. DCI is a substantial but varied market where a range of approaches will be required. When the demand exists then new dedicated DCI systems can be used and when a broader range of functionality is required or more flexibility then metro WDM systems will be more applicable.
Packet-optical convergence
Another hot topic is the continued trend towards packet-optical convergence, which naturally affects the metro market. Again this is a significant discussion that we’ll only be able to look into briefly here.
Packet-optical systems vary hugely in terms of their Ethernet switching capacity, functionality and transport performance. Some systems use centralised terabit-level switching fabrics, but come at a cost in terms of price, physical size, power consumption and transport performance, in areas such as latency, jitter (latency variation over time) and synchronisation support. These systems are often referred to as metro-core packet-optical transport systems (P-OTS) and are better suited towards the core of the network and the interface between metro and long-haul networks.
An alternative approach is a switch-on-a-blade architecture that incorporates the switch functionality on an individual blade, or card, as the name implies. These units combine switching (typically 50 to 250G levels today) with 1G, 10G or 100G ports and all the functionality needed for each port to operate as a long-reach transponder port, i.e. forward error correction, optional OTN framing etc. These have the advantage of being compact and can fit into small access node chassis and offer better transport performance in terms of latency, jitter and synchronisation performance. But of course this is achieved with a smaller switching capacity than a fabric-based switch.
These systems are often called metro-access P-OTS and are better suited in applications found towards the edge of the network with 1G and 10G ports as well as across the metro aggregation portion of the network as the aggregated capacity scales to 10G and 100G. In these locations, space and power are critical and, as mentioned earlier, the requirement for service awareness becomes increasingly important the closer the node is to the edge of the network, which in turn drives the requirements for high transport performance in areas such as latency, jitter and in some cases synchronisation support.
So even within metro systems, the move to packet-optical creates further subdivisions based on the approach to packet transport capabilities. In the ideal world a metro product would offer both in a fully flexible way and without a price/performance penalty if either option wasn’t used.
Can metro rule the world?
If metro systems are expanding into territory that was dominated by long-haul systems only a few years ago, then why can’t these metro networks continue to scale? Distance isn’t the only consideration. Long-haul systems are still a different class of product – they require higher-performance optics than metro systems to span distances of up to 4,000km or sometimes more, and these longer systems will be the early adopters of higher speed 200G/400G optics. Including these technologies in a system would radically change the thermal and power requirements within the chassis, forcing changes that would be highly negative for a system that was also targeting the metro market. Furthermore, long-haul systems require subtly different feature sets as mentioned earlier, and bringing all these into a metro platform would have the same negative effect.
There will always be a clear market for long-haul systems and a separate but partially overlapping market for metro systems. Metro systems themselves can vary as they address the differing parts of the metro network with some highly flexible systems able to span the range of metro-access to metro-aggregation to metro-core networks. The specifics of what constitutes metro-access, metro-aggregation, metro-core, long-haul and ultra-long-haul all vary depending on a person’s specific views and experience. Long live the diversity within the optical networking industry as it benefits us all.
Jon Baldry is head of marketing at Transmode