June 2002

Long Live The King! Migrating Interactive Communications To IP Networks

BY JIM HOURIHAN

The King is dead. Interactive communication rules� again and again and again. For decades, the experts have been telling us that content is king. Whoever controls content, they say, will profit the most from network services. But history disagrees. Interactive communication has always made more money than content distribution. In the nineteenth century, before telephones and e-mail, newspapers accounted for 95 percent of U.S. postal traffic by weight, while ordinary letters generated 85 percent of postal revenues. Alexander Graham Bell thought that people would use his telephones to listen to distant concerts. That didn�t happen. Napster, while a huge overnight success in terms of users, never made any money. Today, data consumes half or more of public network bandwidth, but ordinary telephone calls generate over 80 percent of total earnings. In fact, the telephone industry earns as much in two weeks as the movie industry makes in a year.

So the next time the experts tell you that content is king, tell them that the king is dead. But when it comes to making money, interactive communication � whether it�s person-to-person, person-to-business, or business-to-business � always rules over content.

FOLLOW THE APPS
So what does this mean for the future of networking? Smart network service providers must use their broadband pipes to profit from advances in interactive communication.

Of course, it�s not entirely clear what the next advances will be. The day is coming when instant messaging will likely expand from text-only to voice and video at the click of a mouse. Online shopping will replace static photos with moving pictures and customer assistance based on high-quality interactive voice or video. But most likely, the next big advance will be something that we haven�t yet imagined. And once it arrives, we � or perhaps our children � won�t be able to live without it. We�ll be ready and willing to pay for it by the minute, by the user, by subscription, or by any reasonable combination thereof.

Service providers must be ready, therefore, to support IP-based interactive broadband communication. They must be ready to carry millions of one-on-one high-bandwidth sessions with an unpredictable mix of voice, video, and data. They must be ready to carry the sessions reliably; to convince customers to pay those premium prices. And they must be ready to carry the sessions profitably, with efficient bandwidth use, achievable service level agreements (SLAs), and effective accounting and billing systems.

THE SIP TIDE LIFTS ALL BOATS
Success will require the right network technologies to make these next-generation applications possible. One key technology is SIP (Session Initiation Protocol). SIP is considered by many to be the preferred signaling protocol for next-generation networks, network equipment, application platforms, and applications. A growing number of major wireline operators have already declared their support for SIP. The Third Generation Partnership Project (3GPP) has chosen SIP as the signaling protocol for advanced wireless communication. Cable Labs has included SIP in the PacketCable 1.2 specification. SIP is also the preferred signaling protocol for next-generation voice and video conference bridges, voice mail systems, unified messaging platforms, and more. Microsoft has included SIP in its Windows XP operating system, and leading equipment vendors like Cisco, Nortel, Siemens, Mitel, and others have built SIP into their IP phones. AOL, MSN, Yahoo, Lotus, and others are standardizing on SIMPLE, a SIP-based presence and instant messaging protocol.

SIP has what it takes. Designed by the IETF to set up dynamic interactive sessions over IP networks, SIP meets all the requirements:

SIP is flexible and extensible. SIP supports voice, video, data, and mixed-media sessions. It allows both stateless and stateful operation and easy third-party modification of protocol messages. With SIP, a user could start out with one-on-one instant messaging, click to shift from text to voice, click again to add video, and click again to add another participant. In addition, SIP can be easily extended to include new types of applications and media.

SIP is scalable and efficient. SIP is lightweight and fits comfortably on a wide range of devices, from heavy-duty application servers for unified messaging, presence, conferencing, and the like, to desktop PCs and mobile phones and PDAs. SIP also provides extremely quick call setup.

SIP is designed for IP. Rather than reinvent the wheel, SIP leverages other IP protocols like SDP, MIME, and TRIP. SIP address formats, e.g., sip:[email protected], are familiar to Internet users, while clear-text protocol and HTTP-like error messages make SIP easy to troubleshoot.

The bottom line? SIP is the critical technology that uniquely enables both traditional as well next-generation interactive communications.

NEW NETWORKING REQUIREMENTS
Still, interactive broadband communication is easier said than done. To maximize its value, interactive communication must reach to anyone, anywhere. To paraphrase Metcalfe�s Law: the usefulness, or utility, of interactive communication equals the square of the number of users. New interactive services, therefore, must become as universal as the worldwide telephone system. Today�s consumers and businesses will be satisfied with � and pay money for � nothing less.

But no single IP network stretches far enough or wide enough to support universal connectivity. Most interactive sessions have to traverse multiple networks to reach end-to-end. A presence-activated instant voice call, for example, may cross two enterprise networks, two service provider service networks, and an IP transit network that connects them. Also, consider the need to span business and residential networks or wireline and wireless networks.

Unfortunately, the Internet � the most successful model for multi-network cooperation � lacks certain critical capabilities. Security and traffic control at network boundaries is primitive at best. The quality and accounting mechanisms required to convert services into revenue range from simplistic to non-existent. So before service providers can exploit the interactive communication opportunity, a few technical challenges must be met.

First, a method is needed for determining the best route for each end-to-end session. Ordinary IP routers find the shortest route between endpoints, but the shortest route may not be the best. The shortest route may, for instance, traverse congested low-speed trunks that add too much delay or drop too many packets. Or the shortest route may cost more than a longer path. To find the best route, a number of factors must be considered:

• The availability of paths between the endpoints. How can you get there from here?
• The QoS of each path. Which paths can satisfy the bandwidth, delay, jitter, and packet loss requirements of this session?
• The cost of each path. Which paths furnish the required bandwidth and QoS at the least cost to the service provider?

Since every session is potentially unique and the state of the network is constantly changing, all of these factors must be considered on a per-flow basis. If no path meets the requirements, the session setup request must be rejected. This is especially important on low-bandwidth links like those connecting enterprises to service provider networks.

Once the best route for a given session has been determined, the media flows must be directed along that path. Across every network border, packets must be forwarded in accordance with QoS policies. Since QoS today is a Tower of Babel, QoS information must be translated or mapped into the proper network-specific format (e.g., DiffServ, MPLS, VLAN) at each network boundary. Moreover, service providers offering premium SLAs for voice and video may not be able to cooperate or trust any packet markings coming out of any other network, including the enterprise. They must be able to control QoS assignments for use in their network directly. Session quality must be monitored and reported, for SLA assurance, problem alerting and isolation, and to guide future route selection.

At the same time, traffic must be controlled to conserve and protect network resources. Bandwidth usage must be policed to ensure that customers don�t violate their service agreements. A low-bandwidth 64-Kbps voice call must not be allowed to turn into a 384-Kbps video session without authorization. Session activity must be monitored to detect and remedy faults, e.g., sessions that don�t terminate properly. And of course, per-call accounting records must be generated. Why bother with any of this if you can�t bill for it?

Throughout all of these activities, the security and privacy of both the network and the users must be protected while also satisfying laws like CALEA (Communications Assistance for Law Enforcement Act) regarding electronic surveillance. Access control must be enforced to block unauthorized entrance into any networks. Network topology and addressing must be hidden at both the signaling and the media levels to guard provider assets and supplier/customer information.

LONG LIVE THE KING!
These critical functional requirements are extensive, complex, and require tight integration. Today�s Layer 3 routers, data firewalls, SIP proxies, softswitches, and other network equipment address part of the challenge, yet none provides the basis for successful interactive communication across IP network borders. A successful solution must integrate four functional elements that make the network session-aware:

• Routing policy and constraints;
• Session routing and signaling;
• Media directing and control; and
• QoS monitoring and reporting.

In a session-aware network, the four functional elements interact to share information dynamically. The routing policy and constraints element guides session routing and signaling in route selection. Session routing and signaling tells media directing and control which path to use. Media directing and control moves packets in compliance with route selection, QoS, and security requirements. QoS monitoring and reporting updates routing constraints with information about actual route performance. Together, these elements create the comprehensive solution required to practically and profitably deliver premium interactive communication across IP networks.

Jim Hourihan is vice president, Marketing & Product Management, at Acme Packet. Acme Packet enables network service providers to deliver premium, interactive communications � voice, video, and multimedia sessions � across IP networks. For more information, visit the company online at www.acmepacket.com.

[ Return To The June 2002 Table Of Contents ]