The core erupts: Network intelligence, quality get pushed to the edge

If you've been thinking there isn't much new in the world of Advanced Intelligent Networks, you've been thinking right. But many forces are imposing their wills on this technology and moving it in a new direction - out to the edge.

"The old AIN model needs a kick in the pants, and IP is that kick in the pants," said Jonathan Rosenberg, chief scientist for dynamicsoft, at the IN World Forum in Miami earlier this month.

The emergence of IP as the transport protocol of choice is the impetus for pushing intelligence from the core of the switched network to its edge, even all the way to the customer. This call to change in architecture was a consistent theme throughout this year's forum.

Debates over the best protocol may continue, but most experts agree that IP is the choice going forward. "You can still point to things that IPX does better than IP, but it doesn't matter - IP won. ATM and frame relay are transition protocols that won't be needed in the future. Nobody is writing applications to ATM," said Tom Evslin, chairman and CEO at ITXC.

With the transport issue seemingly settled, service providers and intelligent network equipment vendors can concentrate on taking advantage of IP's offerings.

"IP is an end-to-end model in which the services work at the ends of the network, not the middle like the [public network]," Rosenberg said.

As IP and intelligent network technologies converge, end users will have more control over service creation and quality of service (QOS). The industry is counting on service creation to drive revenue growth, but creating services that far out in the network presents its own concerns, particularly regarding standards, QOS and interoperability.

"To go global, standards are essential for interoperating in a multivendor environment," said Cynthia Corne, senior director of softswitch product development at Level 3 Communications. In a basic softswitch environment at Level 3, at least four protocols are necessary for network components to communicate, not including the SS7 signaling.

And this is just the beginning. "The nice thing about standards is there are so many of them," said Stuart Rosenfield, senior director of marketing for Alcatel.

If - or when - the industry arrives at a set of interface standards, it will set the stage for the next move, which is service creation. "Standards provide a pyramid of knowledge, but more important, they free people to come up with the real value, which is in services," said Michelle Blank, vice president of global marketing for RADVision.

Service creation primarily is done on two levels. End-user service creation is one level where end users can add, change or delete services on the fly. It is enabled through distributed network intelligence (DNI).

Although DNI has become the new mantra for intelligent network proponents, a proper service creation environment (SCE) must exist first. The underlying environment can take four to six months to implement, Rosenfield said. This second level of service creation includes identifying requirements, doing development work and building libraries, adapting different user interfaces and integration with existing platforms such as customer care and billing, testing and training.

"Service creation is not easy, but once you have it, it's very profitable," Rosenfield said.

Once the SCE is set, end users can access it through any device running what will likely be some sort of JAVA application or applet.

"It may take a few leaps, but ultimately, we want to push creation to the customer," said Douglas Tait, intelligent network specialist for Sun Microsystems.

Moving any type of control out of the network core and to the end user makes the ability to measure QOS much more important. Creating the services also is just a beginning.

"You must have the ability to manage the service. If you can't measure it, you can't manage it," Rosenfield said. Service providers and customers must be able to measure the level of service from both ends.

"At any point in the network, I have to have a clue as to the QOS of that traffic," said Shai Herzog, founder and chief technical officer at IPHighway. This will become increasingly true with the expected growth of application service providers, Internet virtual private networks and the push for bandwidth-on-demand.

To ensure end-to-end QOS, devices and software for setting policies also will have to be distributed throughout the network (see figure). Policies for the network core, which handles mostly backbone transport, will need to focus on congestion issues. The aggregation point in the network, where edge and backbone routers and concentrators reside, will need policies for governing speed mismatches and firewall functionality. At the demarc, where a host of disparate access devices vie for bandwidth, policies are needed for marking, shaping, scheduling and, if necessary, dropping traffic.

And this is all for one network. The issues multiply when traversing different carrier networks or going into the black hole of the Internet.

"The Internet is inherently a multivendor environment. You don't control who you hook up with and won't even know what protocol is being used," said Carol Politi, vice president of product management for Ericsson.

So getting network intelligence out of the core and to the edge through multiple networks means a lot will be going on in the intelligent network arena for a long time.