As we look at all the changes taking place on the internet during the past several years, I think we can boil it down to two simple observations. First, the volume of traffic is increasing exponentially. Email, IM, and RSS all mean more connections. Second, each connection is moving a great deal more data, including multimedia, voice, and video.
What does this mean for the future? My guess is that over time IT departments will surrender control of their networks. The communications infrastructure and its management will be outsourced to providers that can offer the kind of always-on, always-secure QoS that the new volume and type of traffic requires.
I spoke with Nortel Networks' Atul Bhatnagar and Peter Carboni for confirmation on what these new networks will require.
Although offering QoS by peeking inside data packets as they speed through the network has been around for some time, this capability has been mostly for smaller networks. But now technologies such as RSS, SIP, and XML for e-commerce require networking and switching that scales to millions of people, says Carboni, Nortel's chief architect and acting CTO.
Take RSS feeds, for example, where XML is being used to sift through data. Imagine an intelligent switch that reads more than just the end-user's IP address. Instead of routing only through layer 3, the switch could route data based on its content as well.
XML, the lingua franca of e-commerce, now carries highly sensitive credit card and financial information and requires different levels of security, says Bhatnagar, Nortel's vice president and general manager of data networks. According to Bhatnagar, these changes require far more scalable, flexible, and intelligent networks.
The ultimate goal is to accommodate these demands by transforming our current patchwork communications infrastructure into a seamless utility, similar to the electric and gas utilities. To that end, there are three key principles emerging from the community of technologists who work on the next generation of the communications infrastructure.
First, to have an always-on network, such as a utility, there can be no single point of failure. "If there is a failure, there has to be either hardware or software that takes over," Bhatnagar says.
Second, we need a more holistic defence system, a so-called layered defence. Security must reside in every subsystem of the infrastructure -- in the campus or core network, edge switches, VPNs, and beyond the four walls of the company.
Finally, as the packets flow through the communications infrastructure there must be the ability to perform deep packet inspection, whether it is data, voice, IP, XML packets, or something not yet developed. The system should be capable of opening the packet and knowing about the data type, as well as the type of user and the type of application, in order to provide end-to-end QoS, Carboni says.
It is a question of matching the user's needs with the application's needs and adapting the system accordingly. For example, for voice you need low latency. For video, even less latency and jitter. Similarly, web browsing should take a low priority compared with e-commerce traffic flowing through, which should get more capacity and resiliency.
IT managers need an always-on network that is secure, with guaranteed performance and visibility into what applications are running and how systems are performing. It won't be long before the network infrastructure outgrows the ability of any one company to do it all.