Frank - I agree that end-to-end ATM is not exactly prevalent right now. Also, you understand the nuts and bolts of these issues a whole lot better than I do. However, I have been forming that opinion based on what I have read from industry articles and also from extensive research that I have done on TeraGlobal Communications (http://www.teraglobal.com). TeraGlobal is proposing to deliver VVD services as a total, bundled solution (hardware, software, services, telecom, maintenance, support) for a fixed monthly fee. Their solution is based on the premise that you need a processor like the G4/AltiVec to handle multiple, concurrent, two-way, real-time data streams and ATM to the desktop in order to deliver these services with the highest possible quality and performance. Take a look at their web site, I think you'll find it worth a few minutes. The company is working closely with Apple and Motorola, among others. Here is an article on ATM vs. IP written by their CTO, Grant Holcomb.
-------------------------------------------------------
ATV vs. IP
When the first telephones were installed, the telegraph industry scoffed at the new technology and labeled it irrelevant. The telegraph empire with all of its capital investments, dependencies, relationships, customers, and large highly skilled labor force could not possibly be wrong or obviated.
ATM is to IP what the telephone was to the telegraph.
First, it is not technically correct to compare ATM to IP. ATM is a transport layer technology and IP is a communication protocol. An ATM data stream consists of "cells" of equal size ? always 53 bytes long. The first 5 bytes of every ATM cell define its destination and purpose. In contrast, an IP data stream consists of variable length packets. Each IP packet has a header, which can be from 20 to 60 bytes in length. An IP packet can be any size from 21 bytes to a maximum of 65,535 bytes. Just like an IP packet gets sent over an Ethernet transport layer in a local network, an IP packet can be divided into ATM cells and transported over a medium that supports ATM. In fact, over a highly distributed network it is now cheaper, faster, and more reliable to send IP over ATM. It is also true that any communication protocol can be seamlessly and simultaneously mixed over an ATM data stream (for example SNA, Novell IPX, Digital?s DECNet, and AppleTalk). This implies the ability to support legacy applications while taking advantage of the real-time nature and quality of service guarantees that ATM offers.
It is very easy to prove the advantages of ATM through an analogy. Imagine if Federal Express moved packages from New York to San Francisco like IP packets are moved over the Internet. You would have all different size, weight, and shaped boxes arriving at unpredictable times. If too many packages arrived at once, the shipment of every other arriving package would be delayed. The receiving center would need a very large, highly skilled and expensive labor force. This is particularly true because every package must be opened and its contents analyzed before a decision on how to proceed can be made. As the packages are moved across the country they would have to stop in every city and each and every package would have to be opened again. The contents of the package must be inspected before the package can proceed to the next city on its journey to San Francisco. By the time the package arrives in San Francisco, you can be assured that all perishable items are beyond their expiration date.
However, imagine if Federal Express moved packages as data can be moved via ATM. Unimaginable efficiency can be achieved. Every single package would be the exact same size, weight, and dimension. Every package would arrive at a scheduled time. On the outside of every package would be a computer readable address label allowing it to be automatically sent to its destination. The path the package would take between New York to San Francisco would be private, secure, non-stop, direct, and always take a defined amount of time.
The "ATM" Federal Express would have a fraction of the labor force, lower overhead, greater profitability, and be able to deliver "Quality of Service" guarantees on package delivery that the "IP" Federal Express would never be able to deliver.
Compare the cost and performance of a high-end ATM switch to a high-end IP router. Moving lots of 53 byte packets on an ATM switch requires a fraction of the computational power, memory, and lines of operating code in comparison to a dramatically more expensive, slower, higher latency, labor intensive router.
The "Holy Grail" of communications is to eliminate the barriers of time, distance, and cost. As long as the industry focuses on the IP standard as the smallest logical element of data transfer these barriers will remain. IP Routers are "dinosaurs" in comparison to ATM switches. Considering the processing power that exists in a desktop computer, let the recipient of the ATM cells determine what to do with the arriving data. The real-time two-way ATM cells that arrive can contain literally anything ? an IP packet stream, live video stream, live audio stream, AppleTalk stream, etc. ? all nearly simultaneously. In the time it takes a desktop computer to place a 1,500 byte IP packet into a memory buffer for processing, 28 ATM cells could be opened feeding a real time video stream, real-time audio stream, file download, and email download nearly simultaneously.
The future of communication is to take ATM to the desktop. This does not obviate IP, it just puts IP in its true place as a data store-and-forward technology. IP is not the provider of next generation real-time two-way communication the incumbent IP monopoly would have the market believe. None of the above is conjecture or theoretical in nature. ATM to the desktop is a reality and being demonstrated now ? by TeraGlobal
By: Grant K Holcomb
Chief Technology Officer |
