April 1, 2004

New Rules for New Times

By Justin J. Junkus, Telephony Editor

Telephony evolution has finally become the voice over Internet protocol (VoIP) revolution. SCTE's Emerging Technology (ET) conference this January, and everything I've been reading since then, convinces me that we have moved past the evolutionary stage. Like it or not, that means we have to learn a whole new set of rules and terminology.

The all-digital network envisioned by the speakers at ET brings all sorts of new network equipment. Players in this triple-play game need to understand where telephony fits in a world of core networks, second generation synchronous optical network (SONET), generalized multiprotocol label switching (GMPLS), routing and service aggregation, just to mention a few technologies.

The VoIP revolution didn't happen overnight. A column I wrote a couple of years ago talked about telephony evolution, and advised developing a migration strategy to get from proven circuit-switched technology to packet-based IP telephony. For companies with an embedded base of circuit-switched telephony, that's still good advice.

After all, circuit switching is a viable technology for most telephony features required by subscribers today. The revenue from those features and service bundles will continue to provide a good rate of return on capital, which has been invested in the associated equipment. However, there are few engineers who would recommend building a greenfield circuit switched telephony offering today.

It's about packets

PacketCable is part of the reason. It has given us a choice of vendors who build to a standard that delivers the triple play. Another, perhaps more compelling, reason is that the entire telephony industry has embraced packet technology. Everyone from business systems managers to network service providers are buying and installing IP-based telephony systems.

With universal acceptance, innovation in one telephony segment tends to propagate to all the others, along with substantial quality improvements and cost reductions. With all segments of telephony speaking packet, even established standards are affected. Interface specifications that used to be defined in terms of electrical signals are moving toward data flows and packet descriptions.

This brings us back to the need to learn new rules and terminology. As mainstream telephony becomes part of an IP-based data world, it also becomes one of many network services that must be supported by a cable operator. Tom Ruvarac, Tellabs manager of cable markets gave an excellent talk at ET that included a summary of where VoIP fits into IP network requirements along with video-on-demand (VOD) and business services. According to Ruvarac, "All these services need guaranteed QoS in the IP backbone, high availability, wire speed, and layer 2 switching combined with layer 3 routing." These simultaneous requirements would spell prohibitive expense without new technologies.

QoS at the wavelength

New technologies and new rules to manage them are, therefore, necessary. Although the data world is a major source of new technology, not everything is packets and frames. Gary Southwell, vice president of marketing for Internet Photonics, pointed out to me that there are a number of cases where QoS can be improved at the physical layer by using wavelength management rather than routing or switching. We'll be talking more about the interaction of protocols, transport and physical layer technologies in future columns. Before we get into that depth, however, it's important to establish some baseline terminology to compare the public switched telephone network (PSTN) with the data networks used to complete VoIP calls. The analogies between PSTN hierarchies and data network hierarchies are not perfect, but they are reference points that will prove useful when we think about marrying technology built for the PSTN to VoIP.

Hierarchy comparisons

Let's review the telephone engineer's view. The PSTN was built as a hierarchy. The local central office connects to a subscriber's phone and is called a class 5 office. It handles call completion for calls within the same service provider and toll network. To get beyond the service provider or reach a subscriber outside the toll area, calls are completed via an interexchange carrier (IXC) over IXC facilities. Access to the IXC is at a point of presence. In the old Bell System, these nonlocal calls were passed to a Class 4 switching office, and from there to the rest of the world via Class 3, 2 and 1 offices.

VoIP also travels across a network hierarchy, but that hierarchy is described differently. Perhaps the most widely used terminology comes from Cisco, which defines a three-layer data network hierarchy consisting of access, distribution and core layers. Unlike the open system interconnection (OSI) model that separates hardware as a separate layer, these three layers mix hardware and software.

The access layer is the customer interface to internetworking resources. This is where a subscriber plugs into the hierarchy. It is analogous to a connection to a Class 5 telephone office. This is where PacketCable's multimedia terminal adapter (MTA) interfaces to the network.

The distribution layer is the communication point between access and core. It provides routing, filtering and wide-area network (WAN) access as required. Some literature refers to the distribution layer as the metro layer, although metro layer functionality may extend into the core.

A rough analogy to PSTN telephony may be that the distribution layer provides an interface between networks similar to the way an old Class 4 switching office bridged Class 5 offices to the rest of the world. Apart from Cisco's terminology, some literature describes edge networks as the interface with core networks. Edge networks may be viewed as a subset of distribution layer networks.

Most agree that core describes the high-speed backbone that interconnects subnetworks. The core switches traffic as fast as possible. Within the core, there are various service networks, such as frame relay, asynchronous transfer mode (ATM), SONET/SDH, or IP/MPLS. What is important is that switching is the primary function done at core, rather than routing, to ensure minimal delay. Interestingly, the first place where telephone hierarchy became data hierarchy was at the core, when service providers discovered that packet transport was far more economical than circuit switching.

All of this background is the foundation to understanding new products and services that are making it possible to build networks that can simultaneously offer carrier-grade VoIP with video and data. I suggest keeping this column on file, because in future months, I will reference the hierarchies when we look deeper into second generation SONET, GMPLS and service separation by wavelength. I don't know about you, but I'm glad to be back on a monthly schedule. We have a lot to talk about. �

Justin J. Junkus is president of KnowledgeLink, Inc. To discuss this topic further, you may email him at jjunkus@knowledgelinkinc.com