By Morteza Ghodrat
In order to mirror SONET/SDH networks, Carrier Ethernet vendors developed proprietary means of providing protection switching and OAM. A new series of standards opens the door to interoperability.
For more than a decade, proponents of Ethernet in the carrier environment have treated the “sub-50 ms” target of SONET/SDH TDM switching as a golden mean for achieving fault-resilient performance. As the first practical carrier-based Ethernet switches arrived on the scene in the late 1990s, many relied on proprietary approaches for failover link switching and operations, administration, and maintenance (OAM) software.
Today, thanks to the work of the IEEE, ITU-T, and the Metro Ethernet Forum (MEF), single-vendor approaches have been supplanted by a suite of standards defining Carrier Ethernet protection switching and OAM. The acceptance of such standards enables service providers and multisite enterprise managers to feel confident in adopting multivendor approaches for Ethernet transport.
Why Ethernet? Why now?
A few service providers still may wonder about the justification for shifting to connectionless, packet-oriented communications in end-to-end public networks, and how such communication networks came to be defined by Ethernet framing at Layer 2 and TCP/IP protocols at Layer 3. After all, Ethernet was developed strictly within the LAN community and was never intended to be a metro or long-haul transport option.
Yet Ethernet has become the default unified transport of choice primarily because of its ubiquity in the LAN. No other protocol from the carrier community—Frame Relay, X.25, ATM, etc.—has come close to displaying the established base realized by Ethernet. And Ethernet showed that, given the availability of adequate bandwidth, it could carry a variety of traffic types. In addition, Ethernet scaled to large sizes and distances far easier than competing transport technologies.
Subsequently, thanks to years of effort by the MEF and the Broadband-IP/MPLS Forum, service profiles for Ethernet have been defined that allow a Layer 2 service defined for the LAN to enter realms such as circuit emulation and isochronous streaming. For example, the MEF worked with the IEEE, Internet Engineering Task Force, and other coalitions to define such services as E-Line, a point-to-point service to supplant private lines; E-LAN, a multipoint-to-multipoint service to emulate LANs and VLANs over a WAN; and E-Tree, a point-to-multipoint IP service used in such topologies as IPTV networks.
Carrier Ethernet began to win hearts and minds in the service provider community in 2001–02, as the transport costs associated with “ruggedized” large Ethernet switches proved significantly smaller than the costs of ATM switches, SONET/SDH add/drop multiplexers, or WDM systems. Ethernet access devices designed for interface at the enterprise are slowly taking the place of T1/T3 access systems.
Nevertheless, service providers have been anxious to see the true fault-tolerant capabilities of SONET/SDH and TDM networks duplicated within Ethernet. This implies more than simply an assurance that failover in a switched Ethernet environment can take place in the type of sub-50-ms windows familiar in the world of SONET/SDH rings. It also means that Ethernet should offer dependability in all the performance fields, including OAM parameters, that have been prevalent in legacy TDM networks.
Initially, platform vendors responded by offering unique hardware-based protection switching methods that used proprietary failover techniques to mimic SONET/SDH bidirectional switched-ring topologies. But such architectures always were unique to each vendor, since the fabric or hub-and-spoke topologies from Ethernet LANs had to be mapped to what previously were ring networks defined specifically for SONET/SDH.
In the 21st century, however, the adoption of MEF service-level standards has required a move to open, multivendor means of providing OAM for Ethernet switched rings. This means moving to interoperable protection-switching methods in hardware and software.
The drive toward interoperability
Why is adding interoperable protection switching important for a service provider? In the same way that SONET/SDH rings must be resilient enough to recover from fiber cuts without service outages, carriers want to feel comfortable that Ethernet topologies, from metro through long haul, display the full redundancy and fault tolerance that would allow mission-critical traffic to be carried in Ethernet frames. When carrier network planners think of Ethernet, the vision often carries negative assumptions of an enterprise switching fabric that is scarcely fault-tolerant enough for the public network. The new ITU, MEF, and IEEE standards can help give carriers a better sense of how 21st-century Ethernet can meet their OAM needs.
Yet protection switching is but one part of the equation. Network synchronization that relies on local clocks outside of a space-based GPS link is critical, as is an open standard for alarms, fault isolation, wireless handoffs, and analysis of node performance. This is why a standards-based foundation is essential for Carrier Ethernet’s success.
The ITU’s Telecommunications working group, working in conjunction with the Alliance for Telecommunications Industry Solutions (traditionally a source for SONET/SDH profiles), has identified the necessary Ethernet OAM features for the public network. They include defect detection, defect notification, defect localization and isolation, system protection, and performance management. The protection goal includes a failover methodology for recovering from faults in less than 50 ms. The OAM goals have been defined in ITU Y.1731 (“OAM functions and mechanisms for Ethernet-based networks”), ITU G.8031/2 (“Ethernet Linear Protection Switching” and “Ethernet Ring Protection Switching”), and IEEE 802.1ag (“Connectivity Fault Management”).
The common areas of Y.1731 and 802.1ag specify standards for continuity checks, loopbacks, link traces, and alarm indication signals. The ITU Y.1731 specifies additional fault management standards for locked conditions, maintenance communications, experimental OAM and vendor-specific OAM, and additional performance management methods to monitor loss and delay of frames.
The ITU G.8031/2 standard has a different goal, coming closer to underlying hardware in defining the protection switch. The standard specifies unidirectional 1+1 switching, and bidirectional 1+1 and 1:1 protection switching. The work leverages the ANSI and ITU efforts in defining path-switched and line-switched rings, but the Ethernet switching paths are deliberately referred to as “transport entities.”
Since Ethernet’s roots are in fabric and hub-and-spoke topologies, any “rings” in a switched Ethernet network may be virtual rather than physical.
G.8031/2 of necessity must also define an automatic protection switching (APS) signaling protocol for point-to-point VLAN-based links. In practice, this work can map into various transport methods, such as PBB-TE or MPLS-TP. For example, the use of bidirectional 1:1 topologies with single-phase APS protocols maps into the IEEE 802.1Qay transport schemes.
Many participants in the standards bodies and coalitions for Carrier Ethernet recognized at an early date that protection-switching protocols had to be augmented with an end-to-end architecture of distributed clocks for network timing. It was not the addition of isochronous services alone within Ethernet networks that drove the requirement for frequency synchronization—low-latency applications like voice over IP can be handled using statistical means in a standard Ethernet environment, given adequate bandwidth.
Instead, the desire to simultaneously handle multiple classes of service, including LAN emulation and circuit emulation, has driven members of MEF, and the advocates of a common Ethernet public backbone, to recognize the long-term necessity of providing physical-layer clock distribution through an Ethernet port. While long-term planners could anticipate the eventual replacement of T1/T3 services, carriers have to assume that any public network can meet the constraints of true TDM support.
The IEEE’s 1588 Synchronous Ethernet working group effort to define sub-microsecond synchronization of real-time clocks began in 2000, and moved to a draft in 2002. Defining an entire suite of Synchronous Ethernet capabilities has been a joint effort between the IEEE, which has worked on the core 1588 Precision Time Protocol, and the ITU, which has examined the problem from two perspectives. The ITU’s G.8262 defines the characteristics of the Synchronous Ethernet clock itself, while G.8264 describes the Ethernet Synchronization Messaging Channel (ESMC) that carries the frequency information within the Ethernet network. The ITU and IEEE efforts also make use of IETF and ANSI protocol development efforts within the network timing protocol and real-time protocol domains.
Converging standards lead to carrier-strength OAM
Chip makers have begun to embed hardware support for all the switching, OAM, protection switching, and timing standards listed above. They believe that such standards must be applied in tandem to deliver to service providers the end-to-end OAM guarantees they have experienced with TDM networks.
Traditionally, Ethernet silicon vendors have focused on packet-processing speeds and packet-inspection methods, of use in the data center, but often of secondary concern in the network backbone. Carrier demands for sub-50-ms failover in the public network drive chip feature sets more toward the support of specific protection-switching profiles per service, in an architecture that enables performance monitoring and quality-of-service marking on a per-service basis.
The value of such features may be self-evident in wireline aggregation but also applies to mobile backhaul devices using Ethernet for transport. Packet-based mobile backhaul, whether based on fiber, copper, or multi-gigahertz radio as a physical medium, is replacing circuit-based backhaul with services based on Ethernet frames. For such applications, network timing, protection switching, and OAM are absolutely essential. The MEF has defined many of its policing and scheduling profiles with backhaul applications in mind, and chip vendors have begun to design their offerings to support the wireless operator’s upgrade to 3G and 4G wireless services.
Whether or not a particular fiber plant is devoted to wireline aggregation of wireless backhaul, carriers will discover as they shift to the 21st-century Ethernet world that network equipment cannot rely on previous generations of TDM hardware, nor on force-fit architectures from the enterprise world. Carrier Ethernet network equipment must be based on unique silicon products tailor-made for the service provider transitioning to the Ethernet/IP world.
Morteza Ghodrat is director, Carrier Ethernet, for Vitesse Semiconductor. He has more than 16 years of engineering and product marketing experience specific to access, metro, core, and optical products.