It's about a network evolution, not revolution
A new approach to ATM networking leverages existing infrastructure while transitioning to Internet Protocol.
CHAD DUNN, WaveSmith Networks Inc.
There is much ado about the promise of an all-Internet Protocol (IP) over optical network, but what about the matter of that multibillion-dollar investment that the regional Bell operating companies (RBOCs) and many other established carriers have made in the area of services that use an ATM transport? While next-generation IP over optical solutions lead the carrier core migration, multiservice networks won't grow around new technologies that don't also steer the billion-dollar revenue streams that today's ATM networks deliver.
That's not to say that the transition to an IP infrastructure can't or won't happen in many carrier networks. But the fact remains that most frame relay services, voice, DSL, and much of today's leased-line growth is bundled over ATM backbones that were built over the last 10 years. Carriers use ATM because it allows multiple services to run over a shared infrastructure. Today's ATM networks are more than a legacy-ATM is still being pressed even further into the role of supporting carriers' mixed services, because its ability to deliver multiservice traffic from the edge to the core continues to make it valuable.
Next-generation solutions don't yet offer a path that draws ATM-based service streams into the new IP core. They use Multiprotocol Label Switching (MPLS) to control the network, which ATM does not understand. For the effort it would take to lift today's ATM-based services directly into new IP core networks, legacy backbones are bypassed more easily than they could be overhauled. Given the state of the capital markets, the dollars needed for an overhaul are not as widely available as they once were. That leaves the carrier in a tight spot. Does it continue to deploy proven services over ATM or does it make the "Hail Mary" leap to a parallel IP infrastructure (Figure 1)?
But service providers won't abandon their installed infrastructure, and in many respects, ATM remains the best choice for delivering service over whatever infrastructure happens to be in place. Service providers need to leverage what ATM offers, using a next-generation platform that can migrate today's multiservice ATM traffic into new IP-based networks.
ATM switched virtual circuits (SVCs) were conceived on the assumption that users would consume ATM services and that voice traffic would eventually be delivered as voice ATM. The entire network was going to be multiservice in the sense that all services would be invented around ATM. This of course never happened.
Instead, ATM was deployed mostly to support frame relay networks and increasingly to provision network access solutions such as DSL, wireless, and leased-line circuits. Companies like Cascade, Newbridge, and Stratacom led this charge and were very successful. In fact, the majority of today's multiservice networks are based on these platforms, which are essentially ATM core switches with multiservice interfaces tying different service elements together. These legacy platforms are based on original ATM-they are stuck on the assumption that ATM should prevail, not that ATM networks need to migrate toward an IP core. That may be true for some networks, but why would a carrier risk being painted into a corner by continuing to deploy these existing platforms?
The service migration could use a new breed of ATM that's designed to carry multiservice distribution traffic from any network edge into an ATM or IP or any other network core. ATM could evolve into an access technology that takes on different flavors depending on the type of access being supported.
ATM still makes sense anywhere that bandwidth gets squeezed, which is what happens when next-generation technologies meet legacy infrastructure. Short of building out a new optical access plant, most of today's end users won't be linked directly to optical/IP core networks. They will access network services using whatever mix of local infrastructure is available, most of which today is ATM-controlled. That is where ATM can be leveraged to deliver next-generation services out to the edge of existing networks.
From the ground up, a new multiservice switching platform should be designed to help legacy backbones integrate new service architectures. This new class of multiservice switch must recognize that ATM is not the end all/be all of transport technologies. Vendors must allow a migration from fully featured ATM switching to fully featured IP networking without mandating a painful paradigm shift for their customers. Modern hardware and software technologies can easily beat the performance threshold of legacy ATM products, helping the solution to scale better and interwork with other systems more easily than today's networks do.
Indeed, there are technologies that will allow the carrier to open up the data plane to allow switching of both ATM and IP such as serial switching fabrics and other bit transparent switching technologies. Unfortunately, that's just the beginning. An open data plane is interesting, but not terribly compelling, as it infers a "virtual overlay" of management and call control. During the long migration from ATM to IP the carrier is forced to deal with differing management, provisioning, and call control methods that increase operational expense and administrative effort. Carriers need more administrative overhead like they need a hole in the head-that is the real migration battle.
The solution is possible with an innovation that opens the service control plane between ATM, IP, and other network technologies. An "open call model" allows each service to control its own calls across the network platform (Figure 2). Instead of trying to map different types of service signaling between ATM and various edge/core technologies, an open call model architecture opens the control fabric to all types of traffic.
The open call model enables a new breed of multiservice switch, in which one side of the connection could be controlled by an IP call control softswitch and the other side of the connection could be MPLS signaled. For example, if voice is going to be carried over data networks (voice over DSL, cable modem, etc.) along with other multiservice backbone traffic feeding IP service cores, carriers may want those voice circuits to be controlled by something that understands their nature, like Media Gateway Control Protocol (MGCP).
The multiservice control plane becomes transparent, open to any flavor of call control-MGCP for voice, PNNI for ATM traffic, or MPLS for IP. These are some of the signaling technologies that could control calls in a multiservice network. Old ATM systems only know PNNI and static provisioning, but a new breed of multiservice ATM would allow different services to cross those borders freely.
Any of today's existing applications could be expanded and optimized for the constantly changing multiservice environment. Virtual-private-network services, content caching, voice gateways, subscriber management-any of these technologies could be controlled natively through an ATM- or IP-based service architecture using an open call model. Entirely new service technologies could also be supported this way, leveraging whatever signaling techniques they require.
An open call model brings the greatest advantage in opening the multiservice architecture and allowing providers to flex their technology and business options. It is both a transitional (migratory) and lasting change, because the architecture doesn't restrict technology choices and it never will. The open call model not only joins old ATM and new IP networks, but it also creates a service environment that allows them to grow closer together, meld with other network technologies, and even transition apart again if needed.
For example, the different services that are mixed on an RBOC's network might be coming from unregulated subsidiaries and the carrier's separate business private-line and public-switched telephone network service entities. Each wants a means to control its own circuits across the network using whatever protocols and provisioning/billing/management systems make the most sense.
The open call model enables providers to control network elements directly and tie them into existing network-management systems. Whereas legacy platforms use a proprietary management interface that requires a gateway to the carrier's network operating system, the open call model allows the carrier's operating system to interact directly. Circuits can be provisioned with flow-through management, using open protocols like CORBA to interface directly with any of the world-class operational support systems.
The flexibility of an open call model helps to ease some of the looming hurdles that await the migration from multiservice ATM to next-generation IP. It creates a more agnostic environment in which services can exercise native control using ATM or any other transport solution.
In the end, the problem carriers are facing is one of relieving the immediate pain without mortgaging the future of their networks. A next-generation multiservice solution should not force service providers to make an abrupt leap from their ATM networks to new IP technology such as MPLS. Providers can instead leverage the best of both worlds using a next-generation multiservice architecture that opens the control plane between technologies and allows service providers to begin migrating legacy services into an IP-based network.
Chad Dunn is director of product management for WaveSmith Networks Inc. He can be reached at email@example.com.