IP convergence: Nolle, 08/10/98

Putting IP in perspective

By Thomas Nolle
Network World, 08/10/98

It's hard to look around the network marketplace in 1998 and not see what appears to be an IP revolution. People who didn't know what networking was five years ago have their own Web sites today, and some experts are forecasting a future in which

Internet connectivity is as ubiquitous as telephony. Maybe that will be so, but we've got a long way to go. U.S. public carrier revenue is nearly $200 billion, and less than $30 billion of the windfall can be attributed in any way to data services.

There's no question that data traffic will have an impact on the public carrier infrastructure. Often, the peak amount of bandwidth used by data applications is 10 to 100 times the peak bandwidth used in a voice call.

The current public network parcels bandwidth out in fixed multiples of 64K bit/sec, and it simply isn't efficient for bursty data traffic. As we move into the 21st century, 80% of carrier profits will come from non-voice services. More data means more inefficiency, so we can expect the public network to change.

But saying that IP will drive a revolution in public infrastructure isn't the same as saying that the public infrastructure will be based on IP. IP can't reserve network resources to ensure application performance is up to snuff. As such, it isn't the ideal mechanism for carrying voice traffic, H.320 video or other forms of leased-line data. If optimum networks are the goal, moving from a strategy that isn't optimal for data to one that isn't optimal for voice is not logical.

Time-division multiplexing (TDM) is going to be replaced by a form of statistical multiplexing that lets applications use bandwidth in a way that is as bursty or as constant as necessary. But that doesn't mean traditional telephony will go away. The type of telephone switching that gives us dial tones in our homes and offices isn't going to pass away for decades, so switches and IP routers will have to share the transport network that replaces TDM.

We already know what that network will be: ATM. Recent ATM announcements by Williams, Sprint and Bell Atlantic show that the facility-based carriers (which, after all, are financial successes) believe ATM is the multiservice architecture of the future.

This isn't to say that users will buy ATM services or premises equipment. ATM will be buried in the innards of public networks much the way Synch-ronous Optical Network (SONET) is buried now. It won't be a new kind of service, but a way of making the telephone services that are critical to users and service providers today coexist with the data services that will be critical to both in the future. IP has won the race as the User-Network Interface (UNI) for non-voice services in the future. That's a worthy victory, but the fact that IP is the best UNI doesn't make it the best infrastructure.

The marriage of analog voice, IP data, ATM transport and dense wave-division multiplexing optics is going to be interesting and exciting, but it will be difficult as the devil unless we accept what's really going on. Accepting IP's limitations as a network foundation doesn't diminish IP's importance as the foundation of all new profit-making applications. Trying to make IP into something it isn't hurts IP and networking for us all.