December 1998

Interim Integrity: Offloading Internet Traffic From The PSTN

BY KEVIN M. MAYER

Confronted with the reality of dramatically rising Internet traffic, and recognizing the growing burden this traffic places on the public switched telephone network (PSTN), many people feel the need to somehow prop up the PSTN, and to do so with all alacrity. However, the most obvious solution, simply adding more ports and more lines and more trunks, is prohibitively expensive. Instead, service providers would, to the extent practicable, escape the punishing economies of traditional telecommunications technology, and take advantage of the capacities and price points offered by data communications technology. Data communications technology seems more willing to observe Moore's Law, or at least a variant of it, whereby improvements in capacity of any given magnitude actually become less expensive, and dramatically so, with time.

To be more specific, the approach favored by service providers, and recommended by Bellcore, is to create a bridge of sorts between the traditional voice network and the newer, more data-oriented networks. This bridge, moreover, is to rely on the traditional voice network's signaling protocols, known as SS7, and the intelligence residing in and expressed by the PSTN, that is, the Intelligent Network, and the specialized databases that help constitute the Intelligent Network.

The first task for the bridge is to relieve the PSTN of congestion. This task, as just mentioned, is of immediate, even urgent, concern. The PSTN simply was not designed to accommodate the demands posed by dial-up Internet connections. The PSTN, with exquisite care, was designed to accommodate ordinary voice calls, which last, on average, no more than a couple of minutes. The average Internet call, however, easily lasts more than 20 minutes. This difference alone would suffice to throw the PSTN into turmoil; however, there is yet another complicating factor. Internet traffic patterns are more erratic and less predictable than traditional voice traffic. And, as is often noted, data traffic is growing much more quickly than traditional voice traffic.

To divert Internet calls from the PSTN, it is necessary for the service provider to distinguish between voice calls and Internet calls as they arrive from subscribers, and to route them accordingly. Voice calls proceed within the PSTN, and Internet calls are shunted to the appropriate data network. The PSTN is spared.

This procedure might appear, at first glance, purely reactive, an attempt to preserve the status quo. Yet it is, in fact, part of a larger, more ambitious plan, one that proceeds, stage by stage, from the reactive to the proactive. Granted, the first stage seems modest enough: avoid catastrophe. The second stage, however, looks forward to increasingly elaborate hybrid networks that combine the best attributes of the traditional network and the ascendant data networks. The third stage, when it finally arrives, may well be the end-to-end, all-purpose data network, based entirely on the Internet protocol.

Still, it may seem odd that the transition to end-to-end IP should start by preserving the PSTN, the very thing end-to-end IP is supposedly destined to replace. There is, however, another way to look at the transition.

At present, voice over IP, or Internet telephony, lacks the quality, reliability, and scalability that we take for granted with the PSTN. Until refinements to Internet telephony eliminate these shortcomings, Internet telephony will be confined, for the most part, to private, managed networks.

Many of these private networks will be interconnected with the PSTN, if only to extend points of access and egress. Thus, early Internet telephony networks will be hybrid networks, spanning IP infrastructure here and PSTN infrastructure there. This being the case, any lack of integrity in the PSTN would weaken any hybrid structure that would depend upon it. The hybrid structures themselves, the intermediaries meant to pave the way towards end-to-end IP, would lack integrity. Thus, solutions that anticipate end-to-end IP, that serve as interim solutions, demand that we preserve the PSTN. Internet telephony, in a sense, depends on interim integrity.

SS7 GATEWAYS
One way to relieve congestion caused by dial-up Internet access is to deploy an SS7 gateway system (Table 1). Such a system usually includes one or more wide area network (WAN) access switches that are used in conjunction with an SS7 gateway.

But before we describe how the SS7 gateway system diverts Internet traffic, let's look at a typical dial-up scenario. The Internet user dials up the Internet Service Provider by way of the local central office. At the local central office, the call may pass through a Class 5 switch (the originating switch), on its way to an intermediary switch (a Class 4 or tandem switch), on its way to yet another central office and another Class 5 switch (the terminating switch), to arrive, finally, at the ISP. Thus, in this example, one dial-up connection demands the attention of three telco switches and occupies two of the trunks connecting these switches.

Deploying an SS7 gateway system could limit the number of telco switches participating in our dial-up scenario to one. (In addition, it could avoid occupying the trunks connecting these switches.) For example, an SS7 gateway connected to the originating Class 5 switch could communicate with the SS7 network to identify the dial-up request as a data call. Then, the SS7 gateway could communicate with the originating Class 5 switch, via the SS7 network, instructing it to divert the call to a WAN access switch. The WAN would be directly connected to the remote access servers of the ISP

Calls to ISPs could be identified based on the called numbers. Then, the calls could be routed to WAN access switches according to one of several techniques, which might include IN office-based triggers, local number portability routing of ISP numbers, IN single-number service, or *XX service code.

A representative example of the SS7 gateway approach is the Ascend Signaling Gateway, one of several such gateways announced this year. The Ascend Signaling Gateway, which works in conjunction with Ascend's MAX TNT WAN Access Concentrator and uses existing SS7 technology and networks to distinguish between voice and data calls and to route them accordingly. In addition, the first release of the Ascend Signaling Gateway converts SS7 signals to enhanced Q.931 messages for call setup, circuit establishment, and call release of data calls between the MAX TNT and the central office switch.

Ascend's plans for its solution resemble those of other vendors. Specifically, Ascend outlines a three-phased approach:

• The first phase is to solve the PSTN congestion problems to offload data traffic from the PSTN to data networks.
• The second phase is to provide voice across data networks. Ascend plans to use its gateway solution to perform call setup, establishment, and termination for voice and data over IP, ATM, and frame relay networks.
• The third phase is to introduce enhanced services to the data network. Ascend plans to accomplish this by adding TCAP (Transaction Capabilities Application Part) functionality to the Ascend Signaling Gateway to provide IN-based services to data networks. A few examples of the services include alternate call routing and local number portability.

The general trend in this approach, as in the approaches of other vendors, is towards increasingly sophisticated integration between voice and data networks. For example, services such as call waiting and call forwarding, which are common in switched networks, may be introduced to data calls via SS7. To date, users of Internet telephony have done without the more sophisticated services available in the public networks. Such sacrifices may soon be unnecessary.

A NEXT-GENERATION SWITCH
An alternative approach to Internet offload is being developed by Castle Networks, a startup that claims its Services Mediation Architecture, a central office platform, will mediate disparate signaling, transport, and management functions to span multiple networks and deliver voice, data, and converged applications. Like other vendors that aim to integrate voice and data networks, Castle has chosen to begin with Internet offload, while anticipating loftier challenges.

While Castle's immediate and ultimate goals sound familiar, the company does outline a different approach towards achieving these goals. Basically, Castle proposes a new sort of switch, one that can communicate with both circuit-switched and packet-based networks. In addition, the software that mediates the switch is designed to translate between commands pertinent to either type of network. This next-generation switch is capable of switching voice calls onto Class 5 switches using SS7, and it offloads Internet access to the ISP using primary rate ISDN.

APPLICATION PROGRAMMING INTERFACES
Several vendors indicate that they are defining application programming interfaces (APIs) that would allow third-party developers to take advantage of telecom and datacom directories, and to create applications that would span multiple networks. And yet, at the same time, the API could shield the developer from the complexities of the SS7 protocol.

CONCLUSION
For many people, the ultimate goal in communications is the complete triumph of IP, from end to end within the public network. And probably just as many people acknowledge that we will most likely approach the all-IP network gradually, and that it may take years, even decades, before we face the interesting question of whether to preserve the vestiges of the old, circuit-switched public network.

In the meantime, SS7 and IN possess an abundance of untapped potential, including the potential to ease the transition to end-to-end IP. So, it seems odd that SS7 and IN should be destined for obsolescence - and that they could even be seen as the instrument of their own obsolescence - just as they appear to be lending sophistication to a still-primitive voice-over-data world. Perhaps, in this interim period, we'll all have ample opportunity to consider and reconsider the relative merits of "stupid" and intelligent networks. Is it possible that, with time, we'll stop thinking of network intelligence in either/or terms? Eventually, we may even have the luxury of accessing intelligence wherever we might think it convenient.

Kevin M. Mayer is the executive editor of CTI magazine. To respond to the author directly, send an e-mail message to kmayer@tmcnet.com.

www.3com.com