January 1999

Internet Telephony Comes of Age

BY STEVE BAECHLE

A technology can be said to have come of age when prospective users stop asking "why" they should implement it and start asking "how." During the past six months, Internet-based telephony has come of age.

Even as standards committees work to prescribe a means of interconnecting gateways from different vendors, the Internet telephony industry is maturing, with providers using it to slash backbone transport costs, often without users knowing calls are going over the Internet. The first applications to boom have been pre- and post-paid calling cards, fax services, and the wholesaling of long-distance and international minutes. In provisioning these cost-saving services, providers install nodes in key countries so that calls made to access their networks are local. With the Internet as the backbone, cost is no longer distance-sensitive.

As the benefits of Internet telephony materialize, early adopters are finding that strategic network design and deployment are key to maximizing them. Technical planning, deployment, and implementation are part of the key decisions carriers and corporations must make at the outset to reap the greatest benefits from Internet telephony.

CHOOSING A STRONG CARRIER BACKBONE
For carriers and ISPs to attract customers on the basis of cost while minimizing their own overhead, they must construct cost-optimized network backbones. Networks should be engineered to deliver reliable, high-quality voice, and they must scale easily and without service disruption. And because deploying IP-based voice services involves interfacing between the packet-based architecture of the Internet and the traditional circuit-switched PSTN, provider networks must interoperate with various telephony and Internet infrastructures in place around the world.

Carriers must first decide where to install gateways, and how much bandwidth to acquire. Once these investment decisions are made, the next steps are choosing an architecture platform, either a PC- or a chassis-based gateway, and devising strategies for fool-proof back-up and least-cost traffic routing. As always, providers must anticipate growth, change, and disaster, and the gateway architecture decision is paramount because it impacts all other issues, as well as costs.

SPEEDING TOWARD RELIABILITY
Imagine picking up the phone and not getting a dial tone? Most users can't. So while packetized voice quality itself has evolved sufficiently to make Internet telephony a viable service platform, carriers and ISPs must also take steps to replicate the trademark reliability of the "voice world."

Reliability and quality are tied together, since providers need even greater control to guarantee fixed levels of performance. As user acceptance of Internet telephony grows, demand for pricing models based on service level agreements (SLAs) will surge. To meet increasing performance pressure from users, carriers must avail themselves of reliability options from the get-go and avoid costly network overhauls that disrupt service and return on investment (ROI) projections.

HOW IT WORKS
For starters, true carrier-grade gateways are essential. The right platform can go a long way in enhancing traffic prioritization and a carrier's ability to create classes of traffic. Gateways designed for voice, as opposed to modified data solutions, can actually complement the reliability facets of the IP cloud, going beyond things like the Resource Reservation Protocol (RSVP) now inching towards standardization. A strong gateway works with the network, increasing reliability by measuring latency and determining if and when to divert traffic to back up systems.

Reliability means never having to say your network crashed. One way to achieve failsafe backup is to maintain a parallel circuit-switched network. Tightly integrating the packet-based network with the circuit-switched world ensures automatic re-routing during connection failures. It also provides the low-cost redundancy needed to provision different classes and types of services. But this setup can be costly for carriers in growth mode if it means buying switches as well as gateways.

To deal with backup cost effectively, some gateway providers have integrated circuit- switching into the gateway. A mark of a true carrier-grade solution, this approach assures providers that, in the event of Internet congestion or service interruption, voice calls are instantly diverted to traditional networks. The gateway/switch solution also enables the least-cost routing capabilities needed to use bandwidth most efficiently.

LOOKING TO THE FUTURE
Network administrators should look down the road and design a system that will last as long as possible with minimal upheaval. For example, carriers have told us that the hard drives on which some PC gateways run typically have MTBFs (mean time between failure) of about two years. This means that, to ensure reliability, providers face having to invest in new platforms every two years whether they need them for new applications or not. If these gateways are scattered around the world, the downtime associated with the failures can mean significant lost revenue.

A true redundancy strategy should reach below the network level to the gateway itself to prevent one snafu from bringing down an entire node or sub-network. In PC-based systems, this means having a primary and a back-up system at each node. Using standalone gateways, carriers can implement a process with a distributed processing architecture built in.

Prioritization and switching should be handled on a per-port basis, adding more redundancy and distributed intelligence. "Hot-swappability" is key, too, so upgrades and changes can be made on the fly, without having to shut down gateways. The extension of this concept is a solid scalability strategy that permits easy growth, and meets profit projections.

Many Internet telephony carriers are willing to revamp their entire infrastructures if they are not scalable enough. And many are interested in migrating to newer, more flexible systems for their gateways. PC- or server-based gateways typically support no more than four T1 or E1 connections per box, and adding circuits becomes costly and space-consuming. Not to mention the fact that the reliability weaknesses of these systems are magnified with each additional box that is added and linked. By contrast, some carrier-class systems can be configured to consolidate 30 or more circuits into one node. Besides facilitating growth, this approach reduces hardware and gateway investment, as well as management challenges and overhead as the network grows.

HOW DOES THIS SOUND?
We've glossed over the fact that voice quality is improving as the industry refines the algorithms for voice compression. From a network design and engineering standpoint, one major step service providers can take to optimize quality is to reduce the number of "hops" voice calls make. This doesn't just mean adding nodes, which gets costly. Carriers can actually reduce the number of hops in certain applications by distributing more intelligence and administration capabilities to the edges of the network.

Pre-paid calling card applications offer a good example of how distributing intelligence across the architecture can improve quality - and cut costs. Today, most Internet telephony carriers' Interactive Voice Response (IVR) systems and billing databases are centrally located. That means every call gets sent to the central site numerous times during each calling card transaction as the IVR requests card and PIN numbers, and destination phone numbers. If the carrier is based in the United States, and the call is being placed from Colombia to London, the central switch is hardly on the way.

Sending IVR prompts back and forth needlessly consumes bandwidth, adding overhead and delay. But worst of all, this system subjects the voice call to more hops and compressions/decompressions than are necessary. To avoid this, carriers can push capabilities such as the IVR out to remote sites. And once again, gateway platforms that support capabilities like these can greatly enhance the perceived quality of a voice conversation.

In interfacing to circuit-switched networks, Internet telephony providers must support the carrier signaling used in each country where nodes will be located. The most common are T1, E1, ISDN PRI, Q931, and multifrequency (MF) signaling. International carriers in particular should also be certain that their architectures support analog connections, as many do not.

In many ways, carriers and enterprises face the same challenges, including integrating IP-voice into existing voice and data networks. Depending on the equipment a company has installed, there are three ways that corporations can implement Voice over IP (VoIP). Voice can be encapsulated into IP packets and then sent over ATM or VSAT, or a traditional backbone service such as frame relay. To date, most companies have opted for this approach in order to leverage the Quality of Service (QoS) and prioritization inherent in carrier services. Alternatively, packets can traverse an IP-based Virtual Private Network (VPN). VPNs are becoming much more popular as gateways continue to expand prioritization and encryption capabilities. Lastly, voice can be blasted across the public Internet. To date, few corporations have taken this approach.

POSITIONING INTERNET TELEPHONY GATEWAYS
The enterprise environment can approach quality and reliability in much the same way as providers, but faces the added implementation issue of where to position the Internet telephony gateway. Since many already have router networks in place, the strategy should be to leverage that infrastructure even as gateways are added. There are two options for gateway placement. The first is to situate the new equipment behind the router, so that voice and data stream into the router separately and are consolidated for transmission across the WAN. The alternative is to have the gateway operating in front of the router so that voice traffic flows directly to the gateway without going through the router, and is encapsulated by the gateway in IP. The router pumps data to the gateway and, again, traffic is consolidated for transmission on the WAN, this time by the gateway.

The benefits of this approach are that gateways are better engineered for voice, and are designed to add more flexible traffic prioritization than traditional routers. Allowing the gateway to serve as the WAN interface provides more granular control of voice quality and prioritization. Today's more robust gateway solutions bring substantial capabilities to the table, employing full suites of QoS techniques including memory management, packet segmentation control, network congestion monitoring, user definable classes of service and more.

Also, when voice traffic must go through the LAN and the router to get to the WAN, it becomes subject to the threats of LAN congestion, router failure, and the additional "hops" required to get onto the WAN. Gateways may also provide voice-oriented dynamic bandwidth management so that, during times of congestion, the access device scales back on the bandwidth allocated. A detailed prioritization system such as this assures that no voice calls ever get dropped, and that all types of traffic get through without any being choked off completely.

The whole issue of where to situate the Internet telephony gateway may finally force the true human "convergence" of the voice guys and the data guys. To the voice managers, these quality and reliability concerns offset the promise of the savings with lost sleep. Yet data managers are reluctant to place anything between the trusty router and the WAN. If companies are to take advantage of the power of packetized voice, interaction and compromise will be needed.

It is only fitting that the Internet, a phenomena that has managed to bring the world together and prompt the convergence of business and technology, will finally blur the line between "telecom" and "IS." Just imagine the possibilities for reinvesting the savings produced by Internet telephony.

Steve Baechle is Vice President of Technical Consulting for Hypercom Network Systems. Hypercom is a leading provider of carrier-class IP telephony gateways and infrastructure technology, and a provider of voice/data integration solutions to large enterprises. For more information, visit the Hypercom Web site at www.hypercom.com.