Datacomm

September 30, 2004
IP/MPLS as a basic architecture in future telecom networks and new packet services

Antti Kankkunen

UNITED STATES -- Since the late 1990s, the communications industry has shared a common vision of a future IP-based network infrastructure, where any service can be accessed over multiple access networks, in any location at any time. While layer 3 of the network converges to IP, layer 2 is dominated by Ethernet technologies, which provide the most economical high-speed connectivity. The industry players almost ubiquitously share the long-term goal, but a multitude of opinions exists on the best way to get from today's network to this future infrastructure.

This article presents a set of migration strategies, which enable this transition from TDM to IP/Ethernet to take place gracefully, optimizing the required network investment patterns and maximizing the profitability of a communications service provider.

Service provider strategies
After the exuberance of the late nineties, the communications industry has returned to the old fundamentals of cost control and profitability. Today, a typical service provider network evolution strategy consists of three main elements: optimize, leverage/revitalize and enable new growth. Different service providers give a different weight to these three elements depending on the individual local circumstances, but most of them execute this type of strategy in one form or another.

Building public network infrastructure requires such large investments that the service providers must exploit the full revenue generation capability of their networks and maximize the lifetime of the existing infrastructure. Two out of the three strategies mentioned above deal with maximizing the return from the existing infrastructure. First, it is absolutely necessary to optimize the use of the current network. Providers must focus on network operations, streamline processes, extend the network infrastructure life whenever possible, and concentrate on core competencies.

After making sure that the existing operations are as optimized as possible, it is necessary to seek ways to leverage the existing infrastructure and the existing services. Leveraging the existing infrastructure means enabling it to support new services. A good example of this is the addition of data capabilities to the existing SDH infrastructure. Leveraging the existing services means capping some legacy platforms and growing existing services with new solutions. These new solutions need to support not only the legacy service, but also new IP/Ethernet services. A good example of this is the deployment of multiservice routers, which can support the existing ATM and FR services, and are also able to support future-oriented IP/Ethernet services.

The third strategy addresses longer-term future oriented investments. As we share the final goal of an all IP-based infrastructure, it is necessary to consider the option of investing into IP optimized platforms. IP optimized platforms support native packet switching and are based on Ethernet and IP switching technologies. The strategy of futuristic IP/Ethernet service offerings has to be carefully balanced with the need to support existing services. Often, the new, pure packet infrastructure needs to be deployed in parallel with the old one, which was originally optimized for TDM services and which has been overlaid with ATM/FR switching networks.

MPLS network architecture
One key characteristic of public network services is the fact that users are very slow to abandon services that have been integrated into their business processes. Thus, a key requirement for the path toward future IP network is the ability to keep on supporting the existing service portfolio, while slowly adding the new IP-based services to the network.

The most popular connectivity services today, in the order of popularity, are Internet access, TDM leased lines, Frame Relay and ATM. In the public network context, it is impossible to consider deploying proprietary solutions or even 'standard' solutions without wide industry support. The single standard with real support for this convergence of services to a single infrastructure is MPLS. ATM was a good convergence standard in the nineties, but falls far short of MPLS when it comes to including the IP/Internet/Ethernet in the set of services to be considered.

However, even MPLS has its limitations. The support for TDM services is not properly standardized yet and economic study reveals that an SDH infrastructure is more cost efficient for producing TDM leased lines than circuit emulation over packet infrastructure, if the SDH infrastructure already exists. If the SDH does not exist yet, a case can be made for pure packet infrastructure and circuit emulation. If SDH already exists, carriers have to wait until TDM represents a small percentage of the total traffic mix before an MPLS infrastructure can be more cost efficient than using the existing SDH network. In this latter case, the justification for moving TDM to packet infrastructure comes from the operating efficiency (opex) savings obtained by retiring the SDH network from service.

The key enabler for the convergence of all packet and cell mode services to a single infrastructure is a new class of network element - the multiservice router. This device supports full-blown IP routing functions and full-blown label switching functions. It combines all of the best properties of both IP routers and ATM switches. The future oriented services that need to be supported are MPLS based layer 3 VPNs (RFC 2547bis standard), MPLS based layer 2 Ethernet VPNs (virtual private LAN Service [VPLS] and virtual private wire service [VPWS]) and IP routing. The migration for existing services is provided with support for native ATM and FR over MPLS, including PNNI (private network to network interface) interworking capabilities. Another set of very important capabilities is the support for service interworking among ATM, FR and Ethernet. Circuit emulation support for enabling the production of TDM services in the same infrastructure will have to be in the medium term roadmap for a successful solution.

The deployment of MPLS starts in the core network. Most carriers in the world are today busy putting this MPLS core network in place. Once the high-speed core network is in place, the next bottleneck, which needs to be opened, is in the access and regional network. Several different technologies can be considered for feeding traffic to the IP/MPLS core network. Table 1 presents some basic properties of these technologies.

Table 1: Possible technologies for feeding traffic to IP/MPLS core networks

Often, the most attractive strategy is to deploy a combination of the last three Ethernet-centric technologies.

Network architecture evolution

Leverage/revitalize: data enable existing SDH infrastructure - Today's transport network is based on SDH. SDH (or SONET) based transport network extends to virtually every telecom exchange in the world. It would be unthinkable to ignore this vast investment on the way to the all-IP network. Fortunately, SDH is currently being re-invented. All future oriented SDH vendors are enhancing their existing solutions with next-generation SDH (NG-SDH) extensions, including higher density interface and switching modules, integrated DWDM or CWDM capabilities, NMS/GMPLS support and, last but not least, by data enabling the SDH platforms. Existing SDH devices are adapted to carry IP/Ethernet payloads with the support of GFP (generic framing procedure) encapsulation, virtual concatenation, link capacity adjustment scheme (LCAS), Ethernet service interfaces and integrated Ethernet switching. The NG-SDH feature set enables carriers to leverage and revitalize the existing SDH infrastructure, which can be used for providing the high speed Ethernet connectivity between customer premises and the IP/MPLS core network.

Enable new growth: MPLS enabled metro Ethernet infrastructure - As traffic volumes grow and the traffic mix is increasingly dominated by packet switched services, it becomes necessary to consider native packet based solutions for access and regional networks. The evolution toward pure packet infrastructure will start from the locations with largest numbers of customers and spread out to less densely populated areas over a long period of time. Incumbent carriers will build the packet infrastructure in parallel with the SDH infrastructure. For a long time, the SDH infrastructure will be used for TDM leased lines and the packet infrastructure will be used for delivering all packet and cell mode services. Finally, TDM services will also be supported from the packet network with circuit emulation.

There are two possible technologies available for packet based access and regional networks, also called metro networks: metro Ethernet with and without an MPLS control plane. Both alternatives can deliver a native packet over fiber architecture, can support statistical multiplexing and are cost efficient at high bandwidths. However, plain Ethernet without an MPLS control plane is challenged in supporting traffic protection, QoS, OAM (operations, administration and maintenance) and traffic engineering in a scalable manner. Supporting the existing services from this single infrastructure also requires MPLS. Therefore, the optimal architecture has one layer of plain Ethernet multiplexing in the access network and after that the traffic is brought to an MPLS enabled metro Ethernet network. Often, the access multiplexing solution based on plain Ethernet can be a DSL solution.

Summary of network evolution
Let's take a step back and look at the fundamental end user needs driving the architecture of the network connectivity layer. The end user needs can be divided into four rough categories: private IP (intranets and IP-based machine to machine communication), public IP (Internet), ICT value-added services and voice services, respectively.

Today's network infrastructure is built on top of an SDH transport network. The most popular services for fulfilling the need for private IP connectivity are TDM leased lines and ATM/FR switched data services. The leased lines are produced in the SDH infrastructure and the ATM/FR switched data services are produced in a FR/ATM overlay network running over the SDH infrastructure. Internet services are produced in an IP infrastructure, which is being MPLS enabled at the moment. The core of the IP network runs increasingly directly over DWDM/fiber layer, whereas the IP network runs on top of ATM/FR or SDH in the access network. Voice services are produced with TDM voice switches, which utilize the SDH transport.

Fig. 1 shows the migration steps that enable the network to evolve to an architecture with an IP/MPLS connectivity layer running either on top of the NG-SDH/DWDM transport layer or directly on top of fiber. All the migration steps are, to some extent, taking place in parallel and not necessarily in the order of the numbering used below.

Migration step 1 is the introduction of NG-SDH, which enables the SDH infrastructure to support Ethernet service interfaces and flexible packet mode connectivity. This can be sold directly to end users in the form of Ethernet private lines or it can be used by the IP/MPLS layer as a more flexible transport service.

Migration step 2 is the introduction of MPLS based L3 VPNs and MPLS based L2 Ethernet VPNs. The current service landscape, which uses leased lines (LL), ATM and FR for interconnecting corporate networks, was created during the 1990s. At that time corporate networks used a multitude of different protocols and LL/FR/ATM were the services best suited for providing corporate connectivity. Since then significant convergence has happened in the corporate networking. Virtually all of the new network installations are based on IP over Ethernet. Over the years, the IP over Ethernet architecture has gained a dominant market share in the end user networks.

This change has created a situation where the LL, FR and ATM services are actually not the optimal way of providing connectivity services to end users. Expensive adaptations are needed to interface the IP over Ethernet corporate networks to LL/FR/ATM based wide area services. The optimal end user services are IP/Ethernet based services, because they are able to interface to the enterprise networks in native format, minimizing both capital expenditure (capex) and opex spending. Depending on customer needs, operators must be able to provide both L3 IP connectivity and L2 Ethernet connectivity. An MPLS infrastructure is uniquely able to provide both from a single network. The dominant approach to L3 VPNs (also called IP VPNs) is called 'RFC 2547bis'. This provides a multipoint-to-multipoint (also called any-to-any) IP connectivity service. There are two different approaches to Ethernet connectivity, which both need to be supported. VPWS (also called Martini) is a point-to-point Ethernet service and VPLS is a multipoint-to-multipoint service. All three IP/Ethernet services are best provided from the IP/MPLS infrastructure.

Migration step 3 is the use of an IP/MPLS infrastructure for the production of ATM and FR services. From the end user's point of view nothing changes. The user is getting the same ATM and FR service as he received before. However, the service provider is producing the service with new multi-service routers, which can support not only ATM/FR services but also IP/Ethernet services. Investing in this type of element is much more future proof than investment in ATM devices, which can only support legacy services.

Migration step 4 introduces new networking infrastructure for the delivery of value-added services. Examples of these devices are content servers and application servers. Migration step 5 enables the use of IP/MPLS as a transport network for voice services, which are still dominantly signaled, using the SS7 framework. However, over time, voice switching will migrate towards the use of SIP and other native IP protocols and voice will simply become another application in the Internet in the migration step 6.

Future vision widely shared by players
A vision of a future IP/Ethernet based communications network is widely shared among communications industry players. The strategy, which enables migration from the current TDM based network into an all IP infrastructure while maintaining service provider profitability, comprises a clever mix of leveraging the existing network and investing in new growth. Two important new technologies help in this migration. NG-SDH extensions enable the existing SDH infrastructure to handle packet traffic in a flexible manner. Multi-service routers enable convergence of all packet and cell mode services to a single IP/MPLS infrastructure.

Multi-service routers also enable service interworking among ATM, FR and Ethernet services, which is an invaluable tool during the migration process. In greenfield deployments, the business case can be made for producing even TDM services in the IP/MPLS network. However, if an existing SDH infrastructure is in place, the migration of TDM services to packet network is delayed until the point in time when the TDM services have become a small part of the overall service mix. The convergence of end user networks to an IP over Ethernet architecture and the emergence of the Internet as the dominant public communications network imply that the future oriented connectivity services are Internet connectivity, MPLS based L3 VPNs and MPLS based L2 Ethernet VPNs.

The IP/MPLS deployment starts in the core network. In the short to medium term, NG-SDH infrastructure is used for providing the high speed Ethernet connectivity between the customer premises and the IP/MPLS core network. Over time, carriers will start deploying pure packet switched access and regional network. These will typically be based on one layer of plain Ethernet multiplexing in the access and MPLS enabled metro Ethernet architecture deeper in the network. The MPLS enabled metro Ethernet infrastructure will initially be deployed in parallel with the SDH infrastructure, which will still be used for TDM services. Metro Ethernet will first be introduced at the geographical areas with highest densities of subscribers. NG-SDH will provide the Ethernet connectivity to less dense populated areas. Over time, metro Ethernet will extend its footprint and finally will have ubiquitous coverage.

(The author is vice president, product strategy and technology and CTO, Tellabs International.)