Dr Andrew K Bjerring and Bill St Arnaud, CANARIE Inc

Research has Role to Play in Internet's Evolution, Part II

By Dr Andrew K Bjerring and Bill St Arnaud, CANARIE Inc

Use of the Internet has exploded to such an extent in the past five years that it will soon be facing some extraordinary challenges, particularly in scaling up to meet service expectations. Universal broadband access, coupled with very high bandwidth applications such as HDTV, lie at the heart of this challenge. If these expectations are to be met, breakthrough innovations may well be required and the research community can once again play its traditional role: part guinea pig, part early adopter, and part explorer of the unknown. Ironically, until fairly recently, the role of the research community in the Internet's development was considered by many to be largely complete.

The early history of the Internet saw the research and university communities in leadership roles. In the US, they were instrumental in building the first Internet backbone - NSFnet - and the Canadian Internet evolved along similar lines, with the creation of CA*net. By the mid 1990s, however, a large number of commercial Internet service providers had been established in both Canada and the US. Conventional wisdom assumed that the commercial sector would become the driving force in developing and deploying new Internet architectures and infrastructure. Both NSFnet and CA*net were disbanded and institutions in both countries began purchasing commodity Internet services commercially.

At around the same time, new research networking initiatives -Internet 2 in the US and CA*net 2 in Canada - were created to serve the special networking needs of the research community. The infrastructures for both networks were based on commercial services, and the related research activity was focused on new technologies and applications, not on innovative Internet infrastructures. With networking communities in other countries following suit, a global "intranet" for the research community evolved, providing high bandwidth connectivity among institutions to support collaborative research and education.

With the building of CA*net 3 in 1998-99, however, the issue of the role of the university and research communities in Internet infrastructure deployment has re-emerged. CA*net 3 was the first nation-wide "Internet Protocol over glass" network in the world, and is still unique among its peer networks. The core network has access to eight wavelengths on a network using 16-wavelength dense wave division multiplexed technology; and additional wavelengths are available on a set of diverse routes across Canada.

Recently, some of the 11 Regional Advanced Networks connected to CA*net 3 have implemented even more advanced technology, and are using dark fibre, less expensive coarse wave division multiplexed technology, and gigabit Ethernet on each separate channel. Within a year or two, the next generation of these initiatives could see even more use of dark fibre and a significant increase in the number of available wavelengths. This would provide opportunities to experiment with innovative Internet architectures based on new wavelength routing protocols.

More generally, the global research "intranet" that has evolved since the early Internet backbones were disbanded provides an ideal environment for exploring innovative networking concepts. It is heterogeneous and worldwide, and among the applications it supports are some that place high demands on network bandwidth. CAN-ARIE is in discussions with the various parties about deploying and sharing wavelengths internationally, and the National Research Council has expressed interest in this concept to support international collaboration.

Within Canada, the environment created by the regional networks linking virtually every research institution in the country provides an ideal opportunity to further explore the application of wavelength routing technology. The goal is a network that would be responsive to requests from distributed research teams for dedicated wavelengths, each with multi-gigabit-per-second capacity, to link researchers to each other and to their shared resources, be they computing power, data sets, or distributed research instrumentation. Clearly, some of these new Internet concepts could wind up providing exactly the right network architecture required by the new class of research grids discussed in the last issue of RE$EARCH MONEY (R$, September 15/00). And once again, the university and research communities are set to play a key role.