CSD 2018
IP Telephony
By Brough Turner
Today’s digital telephony network is optimized for reliable voice communications. However, its audio bandwidth is based on 19th
century microphones, digitized as efficiently as was possible with 1960s transistors, and transported by a centrally managed circuit-switched network. When computing power and bandwidth were precious, this made sense — but not today.
The telephone network is regulated, stable, and relatively slow moving. While a majority of the world’s citizens have heard radio broadcasts and even seen television, more than half of the people in the world have never made a phone call. Now a revolution is about to
overtake traditional telephony. Named “IP telephony,” this technology is based on much more than just the Internet protocol (IP). To understand the potential for human telecommunications 20 years from now, we must look at the underlying trends.
Moore’s Law has held for 15 years of PC evolution. Twenty more years would imply another 10,000 fold improvement, i.e. 600 Gbits of fast memory tightly coupled to a processor array 10,000 times more powerful than today’s Pentium II. More
likely, the open computing platform of 2018 will have moved beyond PCs. Mainframes were dominant for 20 years (1950-1970). Mini-computers had a similar run (1965-1985), only to be overtaken by the personal computer era (1980-2000?). The next stage will be network-connected, open, programmable consumer devices — programmable successors to today’s cellular phone, desk phone, digital camera, and personal digital assistants (PDAs). These new programmable devices will be sold at less than one tenth the price
and several hundred times the volume of today’s PCs. Envision the personal communication device of 2018 by applying a 10,000 fold performance improvement (per hour of battery life) to some combination of a Palm Pilot, cell phone, pager, GPS receiver, and wrist watch.
With this kind of power at the edge, it is no longer necessary or appropriate to build intelligence into the core of the network. In less than 20 years, all communication will migrate to the Internet. Today, Internet congestion is a
problem for voice telephony, but the Internet’s flexibility is unmatched, and IP is ubiquitous. A few levels of differential service (three at most) will be added at the core, but otherwise the Internet will remain a simple network that just routes bits between intelligent edge devices. The simplicity of a pure IP network also means the elimination of the ATM layer and the synchronous optical network (SONET) layer in today’s networks. Why pay the overhead of SONET protection switching? Internet
routing protocols automatically bypass failed equipment.
A more significant change will occur because the current growth rate for bandwidth substantially outstrips the growth rate for computing power. Fiber capacities of 2 terabits per sec are already possible in the laboratory. IP switch/routers are the only remaining bandwidth bottleneck. Here, Moore’s law, new architectures, and massive parallelism come to the rescue.
Incredible bandwidth will appear in the core first. High bandwidth subscriber
access will take longer and will involve more extensive commercial and political battles. But in less than 20 years, wireless technology will break even this bottleneck. Meanwhile, in only 5 years, rapidly growing Web traffic will outpace traditional voice traffic, making voice telephony a niche application.
With a 10,000-fold improvement in processor performance at the cutting edge and incredible bandwidth in the core, what kind of telephony applications become possible? A
Dick Tracy
watch,
under speech control, will be easy. How about 3D-holographic virtual presence like the projected image of Princess Leia in
Star Wars
, but with better quality?
Today we have static holographic images, with a limited field of view, using photographic technology. But digital image technology is overtaking photography. Computer holograms would require separate image calculations for each viewing angle, but a 10,000-fold increase in processing performance is more than adequate! By 2018, I expect to
converse with someone who realistically appears to be sitting across the table from me, even though we are miles apart.
Brough Turner is chief technology officer of Natural MicroSystems Corp., a Framingham, MA-based supplier of open telecommunication platforms. He can be reached at rbt@nmss.com.
Puzzled by a network processing design issue?
