CANARIE aims to usher in radical new design for Internet research networking

CANARIE Inc is developing a bold new initiative that it hopes will maintain the global leadership in high-speed research networking that Canada attained with its optical CA*net3 network. But before the proposal to usher in a radical new design in optical networking can become reality, the advanced network agency must first convince the federal government to underwrite the $50-150 million required to make Canada a global hub for research networking.

Dubbed CA*net4, the concept would move far beyond the traditional telecommunications architecture by moving control out to the edges and the user. It would create a network of thousands of wavelengths using BGP, the basic routing protocol of the Internet. The grid-style architecture would extend beyond Canada and the US and include links to both Europe and Asia, positioning Canada as an attractive centre for routing Internet research traffic. The initiative has been approved by CANARIE's industry-dominated board of directors and policy committee, and discussions with Industry Canada officials are underway on an informal basis.

As the demand for Internet speed and functionality continues to rise, it is becoming increasingly clear that the huge telecommunications networks are running into a serious scaling problem. CANARIE officials are convinced that a completely different approach to the Internet's research, educational and commercial spheres is essential to maintain Canada's global leadership and satisfy the demands of users. Their task over the next year is to consult with stakeholders and extract a commitment from government to pursue the radical new design. The CA*net3 contract is due to expire in 2002, necessitating a strategy to maintain momentum and reassure provincial and municipal partners that federal leadership will continue.

"The routing software will be the new component. The optics is the easy part," says Bill St Arnaud, CANARIE's director of advanced network projects. "We believe that for massive scaling for the future Internet networks and commercial networks, you cannot scale from a top-down architecture. It has to be bottom-up. ... Let's put control at the edge. Instead of having the big carrier control model, we're going to have users at the edge who are in control."

With the launch of all-optical CA*net3, Canada has attracted considerable attention from several nations that are also trying to push the boundaries of research networking. Projects are being conceived or are now underway in the US and elsewhere to leapfrog past the achievements of CA*net3, raising the critical policy question of whether the federal government wants to maintain Canada's global leadership position. CA*net 4 is being conceived, in part, to keep Canada on top, although it also addresses the future needs of the educational and commercial sectors and would include broader range of activities in support of advanced caching, schools and support for provincial and municipal networks.

"With CA*net3, we took the first step of optical networks and our researcher partners in the provinces are taking the next half step by going from a relatively dense wave multiplex system to a relatively less expensive course wave division multiplex system," says Dr Andrew Bjerring, president of CANARIE Inc. "The critical issue is going to be what the (future) architecture is, with the very high demand for Internet services anticipated. We have to come up with an architecture that is quite different from the traditional 'big iron' network, and one which pushes more and more intelligence to the edges. If the intelligence goes to the edge, it opens up the door to a very radically different architecture."

CA*net4 would also break the domin-ance Bell Canada has enjoyed with the first three CANARIE networks. All traditional and Internet-based telecommunications firms would be invited to participate in building the CA*net4 network by linking thousands of new and existing wavelengths to form an intricate grid of optical connections to allow users to bypass the handful of large carriers that currently dominate Internet traffic. While the traditional telcos have indicated tentative support for the new architecture, it also poses a clear threat to their existing business models, not to mention the huge cost of gravitating over from a large monolithic network architecture. Carriers such as Bell have built successful business cases based on the traditional architectural model, which will be thrown to the wind as control moves out to the user.

"We're going to be experimenting with these concepts on research networks but how quickly and which parts of the commercial Internet will pick up on this remains to be seen," says Bjerring. " The traditional telcos have a very significant hurdle in absorbing some of these challenges of scaling their networks into what CA*net4 looks like. The direction that is natural for them to take is to keep a parallel between this scaling-up problem and the scaling-up problem that they've faced through the telephone networks over many decades.... We're the experts on what we can do at the edge, and how we can emphasize the edge technology in such a way as to minimize the pressures of the core of the network and yet allow it to scale. That's the next generation."

QUICK FEDERAL RESPONSE REQUIRED

CANARIE officials are hoping that the federal government will respond positively and quickly to the CA*net4 pitch, with at least an endorsement of the concept if not a pledge of funding. There is concern that if Cabinet and Industry Canada delay their response, the wrong signal will be sent to the provinces and municipalities that are building out their network infrastructures. St Arnaud adds that CANARIE has a contingency plan in the event that federal support is not forthcoming that includes linking up with the US Internet2, effectively surrendering Canada's leadership position.

"The sooner we give reassurance to the community that something else is coming, the more quickly the plans of the provinces will keep moving. They won't get stalled. If there's the slightest hint that the feds are stalling or it gets dropped as a priority, it will mean the provincial initiatives will lose steam," says St Arnaud. " We saying, 'Do you feel it's important that Canada maintain world leadership? Is it important for businesses, for our society, jobs and growth that we become the leading edge?' If they say 'No, it's not important.', well that's their call. It would be fantastic to get a commitment from government because that gives us the ability to negotiate with the fibre guys."

The CA*net4 concept involves the purchase of wavelengths (the different colours of light contained within optical fibre) to create the new grid architecture. St Arnaud contends that buying those wavelengths within North America and across the Atlantic Ocean will be relatively inexpensive, while the biggest cost would be incurred by establishing a link across the Pacific Ocean.

"We think it would be a big advantage for the content and applications people. It's like a co-location facility - you put in your web servers and content where all roads meet. You don't put them off on a side road and right now we're a side road to the US," he says. "If we can get research networks to (go through Canada), we think the commercial networks will follow. So we want to put in a little grease and put in wavelengths. These wavelengths would be part of the (recently formed) International Transit Network Committee."

To get the CA*net 4 campaign moving, CANARIE has helped initiate several small research projects to conduct simulation testing of the routing software required to run CA*net 4. Using the BGP protocol, prototypes of the routing software are being developed and will be first deployed in laboratories between CANARIE and the Communications Research Centre, both in Ottawa. A working paper has also been completed describing the new protocol, to give the research networking community something to work with as the concept evolves. JDS Uniphase Inc is about to ship an optical switch to be used in the prototype network and a handful of small companies are also likely to participate in the project. St Arnaud says the research community must work to overcome its traditional focus on optical engineering and begin to emphasize optical routing software.

If CA*net4 is approved and completed and replaces CA*net 3, the latter will most likely be decommissioned, unlike previous research networks which gravitated over to commercial use. While CA*net3 boasted the highest speeds yet attained by a broadband research network, CA*net4 will focus less on speed and more on the transformation of the network architecture.

"We want to get off the speed thing," says St Arnaud. I always say it's a male thing to be the biggest and fastest but that's not the importance here. The key thing here which will fundamentally change things is moving control to the users. Whether control to the users is terabyte or gigabyte is not the issue."

R$