Asynchronous Transfer Mode (ATM) and Competing Technologies for High-Speed Computer Networking

Transcript of a talk given by Dr. Raj Jain at Korea-US Science and Technology Symposium, Chicago, IL, April 22, 1998.

Slides used in this talk can be found at http://www.cse.wustl.edu/~jain/talks/atm_kr.htm

Jain: In this presentation, we talk about four different topics. We start with what's happening in networking. Then we go into how it will affect our life and then we go into ATM networks and how does they compete with other networking technologies..

One thing I want to say before hand is that please feel free to interrupt me during the talk. If you have a question, comment, agreement or disagreement, you can raise it right then and there. I have promised the chair that I will not take more than 30 minutes. So somehow we'll make it in 30 minutes. I would really appreciate it if you make it a full duplex, well, not really full duplex, a half-duplex communication. Which means that both of us talk.


Communication today is more critical than computing. We have more computing power than what most of us need. Some of those greeting cards that sing when you open them have more computing power than all of the computers in the 1950's. Some of the games that our kids play today have more computing power than some of the supercomputers of the past. So, next time, when you're working on your computer and you're waiting for something to happen, notice what you're waiting for. You're not waiting for computing. You're waiting for I/O. You're waiting for something to come from the network very likely, or something from your disk to flash onto your screen.

If you want to improve your productivity, if you want to reduce the number of minutes and hours and seconds you're wasting sitting down there, what is there to do? Get a faster network for yourself. If you're a company and if you want to increase the productivity of your people, if their time is your money, what should you improve? Your network! And if you are a country and you want to increase the productivity of the country, what should you improve? Your network! This is one of the key messages I want to give --- both to Korea and to the United States. The network is as important for us today as were the roads of the past, as were the highways of the past. The government has spent so much money making sure that we can get from here to there fast. Today getting from here to there means getting there by the network. That's the only way. We can leave this to private businesses, we can leave it to the government or we can do it both ways. I suggest that we do it both ways so that all of us benefit - those of us who can afford and those of us who cannot afford. If our networks are slow our profits will be slow.

Very recently, about six months ago, I went to Korea. Korea is a very high tech country. It should be obvious from this conference and was obvious to me by having worked with Korean companies for quite some time. However, when I visited there, one thing that surprised me was the network. I think Korea can use a faster network. That will really help a lot. Because if the people there, the scientists there are waiting for some information to flash and it takes two minutes, believe it or not, those minutes are very expensive. We need faster communication.

One of the things I wanted to communicate here was that if there are people working for the government in Korea, I think they should invest more money in networking. The same thing applies to those working for government in the United States. The network is the key to productivity.

Internet, as you have heard so many times, is growing exponentially. At the current growth rate the number of hosts on the Internet will be more than the world population very soon. You wonder if that will ever happen. Yes, that will happen because the number of computers that each of us have will be more than one. There will be a computer in your clock, a computer in your TV. There will be a computer in everything that you use. So the number of computers will grow. For these systems to be useful, we need to be able to communicate with them. That means, they have to be somehow connected, one way or the other, to what we today call the Internet.


The impact of networking is similar to that of the stone age. The discovery of the stone made a fundamental difference in the lives of human beings. That is why, it was called the Stone Age. Well, there are so many discoveries that have happened in the past. In the recent past we have had microwave ovens, stereo, VCRs. These have had some impact on our lives. But there are some things that stand out. What stands out? Stone, iron, automobiles, telephones, and electricity. Can you imagine life with electricity today? Well, two years from now they will say the same for networking. Can you imagine life without networking? We won't. Networking is going to become a key part of our life.

In 1994 nine percent of the households in the United States had Internet connections. In 1997, 26 percent and very soon we are going to have 98 percent penetration in the United States. Which means that every home will have Internet connection just like they have a TV. Today, URL, you know www.xxx.whatever-company-name.com is called URL, is more important than your phone number. Why? Because it is the best way to get information. If you want to give information to a millions of people, all you need is one URL and not a million people to answer the phones. Phone is very expensive. If you want to get information in the middle of the night phone is not the way - URL is the way. Right? We looked up in the Good Housekeeping magazine --- it's not a very technical magazine --- but we found 54 URLs in the first 20 pages. The phone numbers are already going away.

Today, e-mail is faster than telegrams of the past. A few years ago if I wanted to communicate with Korea, I would send a telegram. It would get there the next day or probably the same day if I'm lucky. Today, we exchange hundreds of messages every day. Life has become very fast. So what this high-speed networking has done is, it has made a strong impact on our society. It has made an impact on our life in every possible way. Here are a few cartoons about it. These are not in the package, but this will keep you awake at this time.


Today there is no need for us to go anywhere. We can do everything from our homes. There is no need to get out to the office if there is good high-speed connection at home. There is no need to go out for shopping. Most of the grocery stores will supply you groceries over the Internet -- at least they do in Columbus. You can order pizza from the Pizza Hut and they will deliver it to your door. The best entertainment is on the 'Net. The best education is on the 'Net. Everything is becoming virtual. 55 million U.S. workers will be working remote in the United States by the year 2000. So what this means is that as long as we have a high-speed network, we can do things regardless of where we are. And this is fortunately good for countries which are far away.

This cartoon shows what the world will look like in the year 2050. If you have high-speed connection you could be in the mountains or in the City, it wo'nt matter. Before the highways were invented people used to live in the downtown. Once the highways were invented people moved to the suburbs -- because in 30 minutes they could come to the downtown for work and then they have the fresh air back home again. Well, once you have high-speed networking, you can live in the mountains and enjoy the fresh air all the time. This is another message. The message is that it doesn't really matter whether you're in Korea, or China, or India, or Japan. Today you can do almost the same kind of business as if you were in Chicago. These are positive things which are coming up for us - for U.S. businesses as well as the foreign businesses. We need to somehow capitalize on this whole networking thing. That's another reason you need the fastest possible link between the countries -- the networking link -- I mean.

One of the things that many of us dream about - that once we have high technology, it will save us time and that we will have time to have some fun. However, it turns out that that is not happening. We still work the same 40 hours or probably more. But more does get done in the same time. In fact, this has affected research and development.

We have too much growth in one year and as a result we cannot plan too much into the long term. Previously, we used to plan for five years. We used to plan for 10 years. Today, we plan for one year max or one-zero base two years. You know what one-zero base two is? Two years. Because, if you plan any farther than that, you will find yourself in the wrong place. You plan for something and then you will find that a totally different thing has happened. The products are coming out very fast. The products have a very short life time. The Pentium, Pentium-II, Pentium-III, and so on. These products are coming out so fast that you cannot plan fast enough. In terms of networking, we have the examples of ATM and Gigabit Ethernet. Three years ago, no one had heard about Gigabit Ethernet. You plan for one and the other one shows up. As a user we have to be ready for this fast growth. As a provider and a supplier we have to work to get that fast growth. Product development cycles have already reduced. For example, Chrysler has reduced the car design times from six years to two.

Essentially what that means is that the distance between the research and development has decreased. Put another way - the distance between universities and the industries has decreased. If you are in a university, make sure that you work with the industry. Gone are those days when you did your research, published a paper, three years later your paper came out, two years later somebody read it and five or six years later they implemented it. Nowadays, with the kind of research we do, we find it implemented in the products even before the paper gets published. Same thing for the industry. If you are in the industry, make sure you that you work with the researchers very closely.


Now I'm going to change the topic and go on to ATM networks. One of the key things that has happened in networking in the last two years is ATM. What is ATM? ATM networking is a combination - it is the best of both phone network and the data network. Up until a few years ago the phone company had the phone network. There was this Internet, which the phone company had nothing much to do with, it was designed by the computer people. There were two different communities. The phone company designed the phone networks and the computer companies designed the data networks. Well, soon everybody including the phone companies realized that the real market is in transporting data. So the two industries - phone and the computer - got together. The result of the marriage between the phone company and the computer companies was what we call ATM Network. Now this ATM has nothing to do with the banks' ATM machines. Basically ATM stands for Asynchronous Transfer Mode. The meaning will become clearer in the next slide. It takes the best of the data network and the best of the phone network and combines the two.

How are ATM networks different from the phone networks? Well, the current phone networks are all synchronous or periodic. Which means that the clocks of all phone switches in the world are synchronized. Everything runs at a 125-microsecond cycle, which means that if you get a byte to send now, you get to send a byte 125 microseconds later. ATM doesn't use that old method of phone network. It uses a new method like data switching. where you can send anything you want at any time. ATM uses "packet" switching. In a phone network, all rates are multiples of 8 kilobits because of the 125 microsecond cycles. You cannot send less than one bit every cycle or anything less than 8 kilobits - anything that is not a multiple of 8 kilobits. That's why you have link rates of 64 kilobits and its multiples.

Another thing about today's telephone networks is that all high-speed connections are manually set up. With the ATM networks you can dial any speed you want. You can dial a 155 Megabits per second connection and later on you could say, well I want 25 megabits per second, a bit later you could say I want 5 megabits per second, and so on. Thus, you can dial any speed and you can very the speed.


ATM networks are also different from data networks like the present Internet in several different ways. First, ATM networks provide signaling, which allows you to reserve bandwidth in advance. ATM networks are connection oriented while the Internet Protocol - IP - is connectionless. In a connection-oriented network, you declare your needs while setting up the connection. This is helpful for multimedia.

ATM networks provide quality of service. The routing is based on quality of service. A voice connection may get a very different path from a data connection because the two connections have different delay and throughput requirements.

ATM networks use switching. In IP, every packet has an address and is processed individually, which may be slower than switching.

Finally, and most importantly, ATM networks have excellent traffic management. The traffic management technology in ATM is new. Most of the techniques were not known before 1996. Traffic management is required for high-speed networking. IP and most of the other competing networking technologies do not have traffic management. TCP - the transport control protocol - has a traffic management. But is loss based and it is at least 10 years old technology.


The current debate between ATM and IP is very similar to the debate we all have about whether to buy a new house or to fix the old house. Whenever you have something you cannot do with the old house you have two choices. To fix the old house or to buy a new house that has everything you need. The new house is generally very expensive. That is why often the old technology stays and gets modified. The new technology is used by people who are starting fresh. Slow the new technology becomes cheaper and can survive the competition. Well, that is what is happening right now in the Internetworking world. IP is getting modified and ATM is getting cheaper. And both will survive in one sense or the other.

The same argument applies to other competing technologies. One competition to ATM is the Gigabit Ethernet. Gigabit Ethernet has low cost. It is like a old house. It is cheap it does not have new features. In particular, it has no traffic management. At high speeds like this you need traffic management because if you lose something for a few milliseconds, you will lose your several megabits of information.

Similar arguments apply to Frame Relay and SONET. They all have some features missing and they all are working on those features. So that is what is going on right now? The old houses are getting fixed and the new houses are getting built.


This brings me to the end of my talk. Basically I want to give three messages.

The very first message and the most important of all of the messages is that networking is the key to productivity. Networking is the key to your productivity. Networking is the key to your company's productivity. Networking is the key to your country's productivity.

Second message is that the ATM networks combine the best of the phone and the data works.

And the third message is that the key thing that distinguishes ATM from all the past networking technologies is the traffic management. Most of the past networking technologies do not have traffic management.

That concludes our presentation of ATM. If there are a few more minutes left we could take questions.


Questioner: I have two questions. Considering the lack of ATM applications, do you think the ATM will get some kind of wide popularity in the consumer market? That's question one. And question two. Can ATM guarantee all traffic parameters that are usually provided by the IP?

Jain: There are two questions. The first question is: Will ATM really become popular? The ATM popularity is like the stock market. I was working for Digital for fifteen years and every time we made profit the Digital stock would go down. I wondered why. The problem was that they would project higher profit and the actual profits were less than the projection. As a result the stock would go down. The same thing happened to ATM. ATM projections were very high. It was said that it would take over world and it didn't. And nothing does. So that projection has not been met but ATM is really taking over where it is supposed to.

ATM was designed by the phone companies and, therefore, it is performing well in the carrier market. Most of the carriers, AT&T, MCI and most other long distance carriers, are converting over to ATM. Most of the carrier networks are ATM. Yes, ATM is not yet going onto the desktop. On the desktop, there are no movies, no videos and no phones, yet. So the need for quality of service has not been seen. This is not all because of networking. Other factors as well. The video itself needs to be done differently. Local video needs to become widely used first. There are problems even in local video. Right now we don't have that much desktop video. So we don't have a need for some of the features of ATM. This is giving IP the time to fix itself.

Your next question is what about fixing IP. Yes, it can be done and is being done. Two years from now, this may be true for all the technologies we have done in the past, you won't be able to distinguish IP from ATM. What you will have is IP'. You can call it IP' or you can call it ATM'. All the good features of ATM, for example, switching are already being borrowed by the competition. IP is switching today. You don't route IP, you do switch IP.


Questioner: The question is about DAVIC, which is actually more of a video organization. They've changed their direction from ATM to IP.

Jain: Here is the problem for any video manufacturer today. If you want to make money in 1997 IP is the way. There is no ATM and not at least on the desktop. So today if you want make money you have to somehow sell mostly to the old houses. If you start asking that everyone has to move to a new house to watch the video, people are not going to move. This is sort of expected. Yes, I wish that everyone would move to a new house but that is a kind of unrealistic expectation. Most of the investors have to make money in the short term before they can move the resources into the long term. What will happen in the long term is that slowly ATM will come down in price and some people will have ATM. And we will have both ATM and IP. In the past, we have had similar debates of switching vs. routing, bridges vs. routing and so and so forth. And finally you find that most boxes start doing both. So I w

THIS PAPER IS NOT COMPLETE...

Back to Raj Jain's Home Page