Transport equipment is used to manage connectivity between edge/access equipment and metro/aggregation and core networking equipment. Traditionally, the task of such equipment has centered on the need to aggregate and transport multiple, diverse protocols and services. SONET/SDH was the first true transport protocol that could aggregate voice, data, and other common services. As traffic demands grew, the physical-layer channelization method of WDM was developed for optical fiber. Unfortunately SONET/SDH was defined before WDM; as a result, SONET/SDH had no way to manage wavelengths or optical elements such as optical amplifiers and optical add/drop multiplexers.
Optical Transport Network (OTN) was defined to do exactly that. Initially, the only protocol OTN supported was SONET/SDH. But emerging new data protocols created too many stacked layers, and direct mapping of common protocols such as 10 Gigabit Ethernet and 8G/10G Fibre Channel to OTN was defined. Yet this desire for protocol diversity led to a problem in vertical markets such as digital video, resulting in the standard being expanded to accommodate a new ODU-“zero” rate. As a result of these changes, OTN became a complete new digital hierarchy that tried to mirror SONET/SDH and other protocols.
Meanwhile, networking has gone to IP, with Ethernet being the only remaining link layer for IP transport. Traffic-engineered Multi-Protocol Label Switching (MPLS) and MPLS-TP, both of which are present on most modern routers and switches, can provide the connection-oriented circuit management and traffic-engineered channelization required by network operators’ transport organizations. This is particularly true in access networks, which have undergone a dramatic transformation to IP/Ethernet for residential broadband, more recently mobile access, and soon, business cloud services.
So if all traffic in the access network is IP over Ethernet and all the equipment that needs to be interconnected are Carrier Ethernet switches and routers, why do we need all the complex OTN mapping schemes anymore?
OTN as a common WDM transport layer
Early WDM system companies offered unique ways to run multiple protocols, including SONET/SDH, Ethernet, and Fibre Channel, on common equipment by dedicating a wavelength to each protocol. But it soon became apparent that those early WDM systems did not have the same operations, administration, and management (OAM) and fault isolation capabilities as SONET/SDH. The standardization of OTN protocols for SONET/SDH, and the subsequent standardization of multiple mappings for other protocols into a common OTN wrapper, is a leveler that made possible the migration from the SONET/SDH transport network to a multi-protocol transport network. Such networks not only support legacy SONET/SDH connections, but also the more optimized direct mappings of today’s dominant data protocols. Thus, the so-called IP-over-WDM systems are almost always “IP-over-OTN” systems (more precisely, IP-over-Ethernet-over OTN, since IP requires a link layer protocol such as Ethernet). The IP layering terms tend to be used synonymously.
A thin OTN wrapper optimized for cost and power
As the majority of traffic originates, switches, and terminates in IP routers and Carrier Ethernet switches and routers, the time has come to rethink OTN. Why carry the complexity of all the different mapping modes when the real-world traffic has migrated to IP over an Ethernet link layer?
Transport organizations at the network operators have always argued that they need a rigid and connection-oriented TDM-like transport hierarchy like SONET/SDH to effectively manage the transport network. But this argument has become less and less relevant, as most IP traffic is handled in the form of a connection-oriented MPLS label switched protocol and not routed packet by packet anymore. In the MPLS-dominated IP routing world, traffic engineering has become standard through the TE extensions of MPLS. A final claim by network-operator traditionalists, that network management systems should be able to manage these connections rather than a control plane protocol, similarly has become irrelevant with the standardization of the MPLS-TP ("transport protocol”) extensions to MPLS.
So why can’t OTN become the default “thin WAN interface wrapper” for IP Carrier Ethernet switches and routers? It is a wrapper that can be turned on or off in software, without carrying the complexity, cost, and power penalties of all the different OTN mappings that have evolved over time.
Vitesse explicitly advocates the thin wrapper approach: Optimize OTN for IP/Ethernet, and proliferate OTN capability to every WAN Ethernet port on carrier routers and switches so links can easily be managed in an operator network and can easily interface to regional and long-haul WDM transport networks. The result is that a complex OTN mapper/framer becomes an “OTN PHY.” This approach takes cost out of the network where cost is most important – in the access network.
What are the benefits of OTN on CarrierEthernet switches and routers? They are similar to those found when using packet-over-SONET (PoS) and 10 Gigabit Ethernet WAN PHY hardware on the same equipment, with OTN carrying additional benefits as well:
- A seamless interface to transport equipment
- The actual interface to the router can be a managed link with OAM and fault management, rather than a simple 10 Gigabit Ethernet client interface with limited manageability
- Router interfaces can directly source wavelengths into (reconfigurable) optical add/drop multiplexers (OADMs) and filters without requiring separate transponders when equipped with WDM pluggable modules such as WDM XFPs and ZR SFP+ transceivers. This reduces the amount of networking equipment needed in the access and aggregation networks – an important point when mobile data and cloud services are expected to put a tenfold increased demand on network capacity without increasing the cost of the network.
- Additional reach through forward-error-correction in OTN PHYs.
New requirements for IP-over-WDM access networks
There is an urgent need, particularly in the access network, for optimized IP-over-OTN today. All access technologies have migrated to IP/Ethernet or are in the middle of such a conversion (as is the case for mobile access/backhaul). This migration is being undertaken or planned to accommodate the needs for up to 10X higher capacity at equal or lower cost for 4G/LTE networks, and for business services as IT networks migrate to the cloud – be it enterprise or carrier clouds.
There is plenty of talk about OTN in the context of 100G networking and ultra-long-haul networks carrying terabytes per second of traffic. Yet the biggest opportunity for OTN is actually in the access and aggregation network, where the transition to all IP/Ethernet is largely complete and the pressures on reducing network capex and opex are the largest. In such a world, every switch and router port should have the ability to expose IP/Ethernet over an OTN interface with relatively small power and cost differences, thus enabling a radically lower network transport infrastructure.
Mobile access networks in particular have some unique requirements that pre-date the OTN standards and need to be properly taken into account. The most important one is the need for networks to provide timing. Although OTN can provide frequency synchronization fairly easily through synchronous mapping of Ethernet (SyncE), the emerging TD-LTE and LTE-Advanced networks require nanosecond-accurate time-of-day (phase) synchronization with high phase stability (low wander) as well.
Precision Timing Protocol (PTP) IEEE 1588v2 is capable of delivering frequency as well as phase synchronization to packet networks and is starting to be deployed in LTE networks. With proper time stamping of IEEE 1588v2 packets with nanosecond accuracy, such synchronization services can be provided over many network elements using the concepts of Transparent Clocks and Boundary Clocks. But once IEEE 1588v2 packets are wrapped into OTN, the accuracy of the time stamping may be degraded severely. Thus, packet-optimized OTN PHYs need to be able to compensate for mapping delays and delay variations to maintain the frequency and phase distribution capabilities of the IEEE 1588v2 PTP protocol.
If the networking industry adopts the use of a thin OTN PHY for switch and router network-facing interfaces, IP/MPLS/Ethernet transport is available wherever there is OTN management of WDM – the universal packet protocols essentially ride along for free.
Martin Nuss, Ph.D., is vice president, technology and strategy and chief technology officer at Vitesse Semiconductor.