MPLS technology first emerged within the communications industry for IP core networks based on Layer 3 IP router infrastructures, mainly as a mechanism to provide flow-based traffic engineering and more recently to deliver virtual-private-network (VPN) capabilities. Today’s IP/MPLS core packet data infrastructures are ubiquitous, delivering efficient sharing of network resources with guaranteed end-to-end performance.
Reacting to the high growth of data and packet traffic and lower price per bit, network operators are using the capabilities of MPLS to migrate their core data packet networks toward a converged multiservice paradigm, integrating disparate technologies such as ATM, Frame Relay (FR), and Ethernet, while continuing to support IP. Such a common network will realize significant cost savings in network equipment and the simplification of network engineering and operations. Such significant reductions in operational (opex) and capital expenses (capex) are required for profitability. In particular, the domination of opex within the service lifecycle mandates a reduction in the number of network layers and technologies and end-to-end managed services.
Increasingly MPLS is starting to be extended toward the metro- andaccess-network segments to extend the benefits realized in the core and provide a true end-to-end architecture for the delivery of packet data services.
The network requirements for the access and metro segments of the network differ significantly from the core in a number of key areas (see Table). Salient differences include: ring (or hub) physical architectures versus mesh topologies; multiservice versus single-service network platforms; and the number of individual customer flows, which leads to a requirement for aggregation.
Legacy access and metro networks deployed during the 1980s and ’90s to deliver circuit-based voice and leased-line services were constructed from SONET/SDH ring hierarchies but are not the optimal architecture to efficiently deliver emerging bursty packet-based data traffic. The development of new data-centric technologies on SONET/SDH multiservice platforms like generic framing procedure (GFP), virtual concatenation (VCAT), and link capacity adjustment scheme (LCAS) have gone some of the way toward addressing this issue-but their impact is limited. Network operators do, however, gain some key features from existing SONET/SDH metro networks, such as sub-50-msec network restoration and extended performancemonitoring/operations, administration, and maintenance (OAM) tools.
To achieve an end-to-end architecture for the delivery of next-generation IP and Ethernet services to end users, advanced functions that provide efficient and resilient transport, aggregation, and switching of many packet data technologies need to be embedded into access and metro transport networks. Service providers must be able to guarantee performance and service-level agreements (SLAs) for applications as diverse as Internet access, video distribution, and voice over IP. In addition, the network needs to provide scalability, fair resource sharing under network congestion, and the key values of resilience and OAM in both ring and mesh topologies.
Enter MPLS, which acts as a “glue” to converge disparate technologies and traffic types within access and metro networks as well as provide a common technology to integrate with the core packet data network.
MPLS in access and metro networks provides simplicity and support, via encapsulation, of virtually all network protocols. It also has a number of other technical advantages:
• Transport layer independence. MPLS can ride over virtually any transport technology, including SONET/SDH, Ethernet over fiber, and resilient packet ring.
• Interoperability among network layers, architectures, and vendors. MPLS provides a common protocol for access, metro, and core networks, enabling services to be transported end-to-end across many different network segments and architectures with common quality of service (QoS) and OAM capabilities.
• Service identification. Packet classification is accomplished at the edge (ingress) of the network, allowing significant flexibility in identifying specific customer traffic flows or services.
• Service differentiation. An MPLS label-switched path (LSP) can be serviced based on predefined criteria (e.g., class of service) or bandwidth requirements, priority, etc., ensuring end-to-end QoS and SLAs can be achieved.
• End-to-end signaling. The MPLS control plane ensures simple end-to-end provisioning by using standards-based signaling protocols (e.g., LDP and RSVP-TE).
• MPLS is supported by almost all equipment vendors, based on mature standards.
These technical advantages lead to a number of business benefits. For example, MPLS enables reductions in capex and opex. The use of MPLS as a service multiplexing scheme in the access, metro, and core of the network provides an efficient multiservice network and operations environment, eliminating the need for multiple separate data networks. End-to-end service control can be achieved, even across the network core.
MPLS also addresses the scalability issue of access- and metro-network equipment using label stacking, which is a means of aggregating flows that share a common forwarding path. This capability inherent to MPLS greatly reduces the amount of state information that MPLS devices must maintain. This feature not only enables networks to scale, but also isolates different customer flows to improve security and simplify network maintenance and day-to-day operational tasks.
Finally, MPLS is viewed by many as the most logical technology to underpin a portfolio of next generation multimedia and mission-critical data services over a flexible multipoint-to-multipoint traffic-engineered network. Service providers are depending on these new services to drive future revenue growth as revenues derived from voice continue to erode.
MPLS classifies data packets at the edge of the network dependent on predefined criteria. Flows of packets with common forwarding criteria (e.g., the same class of service) are identified by attaching a label to that flow of packets to create an LSP. The packets are then forwarded through the network dependent only on the label. The MPLS LSP can be manipulated across the network by using the label as identification regardless of the underlying transport technology.
The Internet Engineering Task Force (IETF) has standardized Pseudowire Emulation (PWE3), or “pseudowires,” based on an evolution of its draft “Martini,” that provides a framework to multiplex and transport multiple Layer 2 technologies such as HDLC, FR, ATM, or Ethernet, as well as circuit emulation services (CES) for TDM, and encapsulate (or tunnel) them over an IP/MPLS network.
Combining MPLS pseudowires with enhanced techniques such as GFP and VCAT can provide an excellent mechanism through which to evolve SONET/SDH networks, converging the packet and circuit networks together (see Figure 1). This approach delivers high bandwidth efficiency by statistical multiplexing of packet traffic, along with traffic engineering for service guarantees. At the same time, it retains all the benefits of SONET/SDH technology, such as carrier class resilience and performance monitoring (OAM). The IETF and other standards bodies are enhancing MPLS to deliver OAM capabilities and end-to-end QoS and, most critical to MPLS deployment in the access and metro networks, extending the reach of Layer 2 services over an MPLS backbone.
However, MPLS must be extended to include features that are critical within the optical metro and access domains to support this transition from TDM-based SONET/SDH rings to MPLS-centric packet networks. Considering the ring-centric architectures found within access and metro networks, the ability to provide packet-based ring protection and dual node interconnect is critical, as is extending existing OAM tools to MPLS and the services it carries.
Emerging packet data services based on metro Ethernet have very different characteristics from legacy circuit services, which are simple point-to-point bandwidth “pipes.” Typical packet services attributes are characterized by parameters such as bandwidth profiles (committed information rate, excess information rate, committed burst size, excess burst size), QoS profile (defined by delay, jitter, packet loss, etc.,), and point-to-point or multipoint-to-multipoint in nature.
MPLS pseudowires provide support for the provisioning of multiple “virtual” topology options, allowing the network to be designed per customer requirements. Customer services can be point-to-point (transparent leased-line replacement), point-to-multipoint, or multipoint-to-multipoint, transparent to the customer’s network topology, protocol, and addressing structure.
The network operator can build a layered network based on SONET, WDM, or even Ethernet over fiber, which provides resilient transport, and on top of that, create an MPLS layer that provides statistical multiplexing with deterministic performance. Finally, a service layer can be overlayed to deliver standardized services that meet individual SLAs (see Figure 2).
Using MPLS provides the ability to engineer variable and sustainable QoS definitions, with several classes of service (e.g., guaranteed, regulated [overbooked], and best effort) per customer, which can then be applied to network applications with differing transmission requirements. Pseudowires can be defined that are customer-specific and carry only the traffic of that customer, providing a logically segregated “elastic pipe” that delivers QoS guarantees (e.g., controlled delay, jitter, and packet loss), as well as enforceable and verifiable SLAs for that customer flow.
An example of such flexible services compliant with the Metro Ethernet Forum’s E-LAN specification is a Layer 2 Ethernet VPN, based on the IETF’s VPLS architecture. Such a VPN provides multipoint “transparent LAN” service between multiple enterprise locations through Ethernet media access control switching (see Figure 3).
Although Ethernet services are often thought of as delivered within a metro area, through the use of MPLS they can be extended from one metro area to another to achieve expanded Layer 2 connectivity. The use of MPLS in both access/metro and the core thus enables an end-to-end traffic-engineered MPLS connection with common QoS and OAM mechanisms, from the access network to the core and beyond (see Figure 4).
The explosive growth in packet data services is rapidly driving MPLS, once limited to core-network routers, into access and metro networks. MPLS is clearly emerging as the leading technology to converge multiple data traffic types, including ATM, FR, and Ethernet, onto a single, flexible, efficient network. Such convergence will provide an end-to-end service management capability and effective aggregation mechanism optimized for the specific needs of access and metro infrastructure.
Service delivery is an edge-to-edge requirement. The ability to encapsulate a customer’s traffic in an MPLS pseudowire as it enters the access network and transport it across the core and out again over another access network while retaining that encapsulation has huge advantages. These advantages include customer traffic flow segregation and security, the ability to engineer end-to-end QoS and SLAs, and OAM and performance monitoring.
This migration of MPLS into access and metro networks will reshape the infrastructure in the same way it has done within the IP core, creating a dramatic impact on packet data services and overall service lifecycle costs.
Colin Evansis director of marketing and business development at Native Networks (Coventry, UK) and a member of the board of the Metro Ethernet Forum.