|
Articles Tips for the Telecosm Traveller Networks come in all shapes and sizes and they do all kinds of things, but I think the most important network is the human one. Before I start this presentation, I'd like to give you some background and perspective. I have four children, aged 29, 28, 20, and 14. My two eldest daughters remember black and white TV and no VHS; my son, 20, can only remember color TV and VHS; and my 14?year?old cannot imagine life without absolutely everything, and says things like, "Did you play this game when you were my age?" And it's Doom he's talking about. I said, "Well, no, we didn't have computers." He said, "What did you do?" The growth of our technology and networks has been astonishing across only a few generations. My attitude to science, technology and business is pretty much like my attitude to sex: you can read the words, look at the photographs, you can even watch the movies, but until you've had hands?on experience, you really don't get it. I started building companies as a BT employee, and four years ago I got permission to start building companies with my own time and money. I have to tell you that losing my own money was an entirely different sensation from losing BT's. We are as a species such limited and linear thinkers. We don't cope well with complexity and multiple dimensions. Once we get beyond three or four, we're in trouble. Networking and networks might just change this, and I want to explore that possibility for a moment. We are moving from a world of concentrated skill and capability to a world of distributed ignorance. What is wonderful is that distributed ignorance wins. For example we don't have to know anything about the technology of computing, printing or telephone to benefit from them in our everyday lives. Moreover, being connected by multiple networks is almost magical in the way it gives all of us exponential abilities. Most commentators quote the connectivity of modern networks as growing with N^2 the number of participants, but I think it is significantly more. A broadcast network of one transmitter and N receivers has at best an information flow and connectivity of N/2. But of course, not all nodes are attended all the time - i.e. we don't listen to the radio or watch TV 100% of our lives and so the actual outcome is < In the case of a telephone network of N customers, the total connectivity is N (N-1)/2. But again not every telephone call is answered and every telephone is not used equally and the result is an information flow or connectivity that is >1 we get a connectivity and information flow that grows as N^2. Now for the Internet the numbers are much more impressive. If we all have just 300 people in our address books, then in only 3 clicks we have (in theory) connected to every human on the planet. This leads to a model of 'N' nodes stacked 'n' deep, resulting in an information flow/connectivity It's very powerful the way we have built up our technology. People often draw an exponential riser. In fact it is a series of logistic curves sitting one on top of the other, where each successive technology does dramatically more in a shorter time at a much lower cost. The true scale is thousands of times more with each succession of technology - thermionic tube, to transistor, to integrated circuit. The biggest crime we commit in presenting our technological progress is to use a logarithmic scale, but we have no option because it's the only way we can cram it in the space available. But it creates the illusion of a linear progression when it is a far more dramatic exponential rise. For the younger members of the audience I should explain that the thermionic tube was like a transistor with a light inside so you could find it in the dark! It was the device that powered up World War II and got the electronics industry going. In about the 1960s, the transistor came in. It did thousands of times more in a shorter period. It was rapidly followed by integrated circuits, and the same thing happened: each successive generation of integrated circuit did much more for us in a much shorter time. Over 50 years of capability doubling every year (or thereabouts) and 2^50 is a huge number! Anyone who knows anything about chaos and non? linear systems can guess what happens. This rate of progress builds up to a cataclysm, which looks certain to arrive around 2015. It's called Moore's Wall - the end of the silicon revolution. Does everything stop? I don't think so. Because I think there are far more technologies beyond Moore's Wall than we have seen hitherto. So I am not worried about 2015 and beyond, but I am worried about the transition because our historical mode of sorting out problems of sudden technological change, like the change from the longbow to the crossbow, is to organize a few more wars. I also have another worry, well a complaint really. I was born too soon; this technology is going to be exciting stuff. I've got a little movie for you that I thought was about the best way of showing you what technology can do. I'm going to let it run. Okay. Here we go. [Presentation.] Professor Cochrane: That was two kids, a camcorder and a PowerMac. Can you imagine, two kids with a camcorder and a PowerMac working in a bedroom made that movie? Interestingly, they have to sell it on VHS tape, because there isn't the bandwidth to ship that movie with sufficient quality over the Internet. The most efficient means of data transmission that we have today is FrizbeeNet - CDs and the back of a truck is the best network we have. For the first broadcast and telephone networks we can talk about the edge of the network. For the Internet there is no edge. Witness the number of times you receive a virus that you have actually e?mailed to someone else, and they e?mail it back because you're in each other's address book. Now consider of our linear way of thinking. I've been doing this experiment with lots of audiences. How many people understand what an exponential function is? Just stick your hands up. Quite good. How many people can tell me what 2 to the power 10 is? Very few! How many people understand that they have to pay interest on their mortgage? So you do understand compound interest. But you don't link the two together. If we invest $1 for 10 years at a 10% interest rate, we're going to finish with about $2.60. Invest $1 for 10 years at a 100% rate; we finish with $1,024. The majority of people not understanding this is one component of why we're in a little bit of trouble. Let's try another approach: If I pay you $1 today, $2 tomorrow, $4 the day after, how much will I pay you in 10 days? Answer: $1,024. How much in 20 days? $1,000,000. In 30 days? $1,000,000,000. That's counterintuitive. But it's what the technology does. My laptop will be a 1,000x more powerful in another 10 years, 1,000,000x more powerful in 20, and 1,000,000,000x more in 30 years. Its kind of fun that we don't understand this. By the way, technologists don't get it either. We think we do, but it always takes us by surprise. It's a like a tsunami. You sit on the beach, and by the time you see that a wave is a threat, it's too late to run. That's what this technology does. When I think of networks now, I think of everything from the biological, right through to our fiber optic systems. But everything is also physically moving - i.e. mobile. And, interestingly, we're getting time, frequency and wavelength multiplexing at far end in the long?line networks, and a PCB and a back-frame connection level. I tried to see if I could rationalize networks by the mechanisms we use to discriminate individual signals. In broadcast we use frequency and wavelength as the selection mechanism for each channel. (By the way, frequency and wavelength do not exist; they are just mathematical abstractions on time and space). The telephone network uses time and space; Ethernet just uses time. Local area networks employ space and time and so on. So we have got to simultaneously exploit space, time, frequency and all wavelengths if we are to realize a single universal network to replace all we currently have. Axiom 1: Hierarchies don't work. Let me give you an example. In the old telephone network, when everything was bolted down and static, we would make three or four phone calls per day for three or four minutes on average. Now we have a world where we all carry mobile phones, and coffee is a strange attractor and brings down networks at least twice a day. When coffee arrives at a large gathering such as a conference, then cell phones come out en mass, and the network can't cope with the sudden demand overload. Similarly, an accident on a freeway will see the number of telephone calls go from one every 10km to a couple of thousand in 4 km. That brings down that cell. So it's clear that the aggregation we enjoyed with concatenation, with hierarchical, telephone random?type networks is okay but in this future world it will not work. Everything has to be extremely low and flat if we are going to win. We also have to invoke a degree of self?organization. Now, some new numbers. Averages are more or less useless when discussing network design and performance. On the telephone network of old, the peak?to?mean ratio, that is the maximum traffic to the average traffic, was 3 or 4:1. On the mobile network it's about 50:1. On the Internet it very often exceeds 1000:1. That is like everybody in San Francisco coming with me at 3:30 to the airport to get a flight back to the UK. That's the scale of what we're trying to engineer. The only way that we can cope with this is to throw bandwidth at the problem. The good news is that bandwidth is available in abundance. You wouldn't think so, but it is. There are now commercial systems available with 160 wavelengths at 10Gbit/s - and soon they will be upgraded to 40Gbit/s. The time between demonstrator, idea and reality is getting shorter. When we look at networks, it's nobody's fault, it's historical but we have got the bulk of the risk and the investment in the local loop. The local loop kills you. The other elements are insignificant in the number of people, the amount of equipment but they have to be attacked and made much slimmer, leaner and meaner, to get full advantage. It is obvious that we should go for network leniency and strip out as many levels of hierarchy and technology as possible. As a general scaling factor - a halving of the amount of electronics in a network will reduce the people requirements to about 1/4. This non-linear relationship has an incredible impact on operations and economy. When we went all digital, we did so for one reason only: it was more economic to do switching and transmission in all digital global network. The impact of that was to see the introduction of digital services as the rise of the PC created increasing demands. We are now heading to a world where the cost of network control and management is going to fall because we're actually going back to analog as we go to WDM. It's a funny kind of analog. We have got wavelengths that we can modulate as we see fit. In 1986, I came to the conclusion you didn't need switches in the network because there was such a huge bandwidth capacity using multiple wavelengths on optical fibers. So huge is this capability it will ultimately allow everything to be interconnected in a single nation. The phone companies don't like this story because it reduces their network to a water pipe. And that's where it's headed. My guess is that this realization is probably another 20 years away. If you strip out as many of the components and people you can, networks become increasingly reliable. In the networks we have this little anecdote: If you have a switch or repeater station, you employ a man and a dog. The man feeds the dog; the dog keeps the man off the equipment! About 50% of all network failures are down to the people working in a network, they can't resist playing and they make mistakes. Networks should be limited by the speed of light, not the speed of the equipment. We're not going to get anything remotely like real-time services over IP unless we restrict the number of hops to about three, from one side of the planet to the other. More than about three will kill the performance. Now, progress. Everything is improving except the last mile. There is no wide bandwidth anywhere on this planet. Wide bandwidth isn't 1.5Mbit/s, it really isn't. ADSL could have only come after the diseased mind of a Telco - it is fundamentally a stupid idea. The world is not asymmetric. When you get bandwidth you want to push it both ways, you want to exercise it. My son suddenly recently realized that I could afford a TByte server in my home. But terabyte...what would we do with it? So I installed the first 280MBytes. I have a 100Mbit/s local area network and an 11Mbit/s WLAN. Unfortunately it goes out into the real world through a 56K modem. Can you imagine what my 14?year?old says? You can go around my building lot with your laptop and be online at 11Mbit/s. But my son says, "Dad, when are we gonna get some serious bandwidth around here? It really sucks." And I actually think he's right. Imagine, we are just one switch away from infinite bandwidth; but I can't get at it, the phone company is in the way. That local loop is killing me. Interestingly, our ability to produce bandwidth is growing faster than any other commodity and yet we invest all our money in software. It's a bit like panicking over anthrax when we should be worrying about influenza, which will kill tens of thousands of people a year. Anthrax might kill four or five, so let's put all the money into that. We do that as a species habitually - we invest in all the wrong things Now all this is important because the leverage of technology saw the PC industry hit a $100Bn in 10 years, the Internet did the same in 4, and mobile has done it in 3. We (humans) are now the number-one constraint in the delay function of creating wealth. It isn't the technology - it's us. Our ability to subsume, use, and employ technology is the number one killer. We have saved trillions of dollars in businesses across the world by adopting technology and specifically exploiting the Internet. Sadly, this failed to realize the full potential because there is no bandwidth. You can't have a dot.com world when people can't get access. And nobody seems to have picked this up. Yes, there was stupidity. Yes, there was greed. I mean, I was at a venture capital conference last year and jokingly and said, "I want to launch a new company, the company is BuyYourDogFoodFrom Me.com and I'll give you the dog for free." It got three offers of funding. What is interesting is that peer?to?peer, or customer?to?customer is growing like topsy because the music industry stupidly closed down Napster. Now there are 40 million Napster derivatives for them to contend with. They are really going to suffer. Sony just launched content? controlled CDs. You put it in your hi?fi; it works. Put it in your car, it works. When you put it in your PC, it won't work. They spent millions on this. So what do the kids do? Stick it in a hi?fi, feed the output into a PC and re-digitize. It only takes a few minutes. Think about that. It's cost them millions and the kids have defeated them for a few pence. I think we've just had an inkling of what is possible. If two kids can create a movie (www.404themovie.com) like the one I showed, is Hollywood going to employ them? I think it's possible we're going to see a new business where the consumers start to feed business. But it's going to need networks. It can't happen without networks. We should now remember that mobile overlays the whole thing. In Europe now you can buy a Diet Coke using a mobile phone. It's starting to become a security device, it's becoming a wallet, and obviously it's a portal to screens. It's going to change the way we relate to technology. Will 3G happen? Sometimes the human race does something really stupid. A published report in Europe said governments have ripped over $200 billion out of the industry and as a result, there are a half a million people unemployed now across the industry. We have to find a solution. Here is mine: We share the infrastructure; we merge operations. The EU is going to have to plug some money back in, but let's use 3G to wipe out the last mile. Another bit of stupidity. Because politicians can't get their brains around spread spectrum, they've been carving the spectrum up instead of leaving it alone. Carving it up like it was TV channels cripples spread spectrum A common perspective with 3G is you're going to get up to 2Mbit/s. What is the reality? These are the numbers. 2Mbit/s is the amount of bandwidth delivered to the cell site. If you're real close to the base station and nobody else is making a phone call, you might get 2Mbit/s. I don't think you're going to get that, or very little of it. What you're going to get is 64 kilobits. It will be wire-line quality. But curiously, that data rate delivered to every customer for access to the Internet, gives a higher aggregate data access across most countries. In the USA today over 90% of the people have to use a dial?up modem to get Internet access. In Europe it's 95%. The rest are on the motley collection of cable modems and ADSL. Here's the perversity of ADSL technology: where are the high net-worth individuals? The answer: On the periphery of a town. Where does an ADSL or DSL reach? The periphery of the town! So the people who need it can't get it. It's inverse to what we need for the creation of wealth. There's exponential growth everywhere in the mobile world, and in particular SMS. In the UK 12Bn SMS (text messages) are sent every month for a mere 60?million population. In the UK pay-as-you-go mobiles are not only popular, they are the prime mover for growth. You can buy a mobile phone like you buy Corn Flakes. The first Christmas this was the case, over 500,000 were sold in two weeks. The second Christmas, 2,000,000. Third Christmas, 4,000,000. And last Christmas, 6,000,000 were sold in one month in the UK alone. Because they cost $60, they're not only a great Christmas present, there is no contract and they work anywhere on the planet. What's the key constraint to progress? Well, the carrier's are the bottleneck. We can do almost anything but we can't get past their stranglehold on bandwidth. So if Telco sold cars, this is what they'd advertise and this is what they'd promise and this is what they'd deliver. Are they going to deliver? No. They don't want to, they don't have to, and they can't anyway. Why not? Long-lines experience said that when you put fiber into long lines, you get rid of 90 percent of the people. Can you imagine the guys running the local loop saying, "Hey, we're going to vote for early retirement"? If you put fiber into the local loop, you get rid of 90 percent of the people out there. Nobody's going to do that. This has happened before. It happened in the 1940s when TV networks rolled out and it was not cost effective to put TV into every small community. So what did people do? They put up their own towers, amateurs wired up the place with cable and it became community antenna TV. Then it became big enough to become cable TV and has become a major industry. And now we stand on the verge of a similar situation with the possibility of (parasitic networks -autonomous WLANs) being able to do the same. Here is the current state of play with WLAN technologies. Who needs a Cableco or Telco? Could we set up our own network in our community and get it to work? The answer is yes. Last year I came to this country and for a whole week I never used a telephone. Everywhere I went I was able to plug my laptop into a local area network. In the hotel it was free or $9 a day, in the airport it was available at Laptop Lane and in the companies everybody let me plug in. This year, for the first time I came to this country and I came for a whole week and, total fluke I'm sure, but every coffee shop I went to had a WLAN, the hotel had one and the company I was in, had one. It was magic. You cannot imagine how much work you can do when you get wideband access and you're on the road. So let's just play a mind game. Let's think of any institution over here - it could be the local school, local gas station, local superstore or anybody else who's got fiber. They put up an antenna with an 802.11 WLAN. You all go to the store and for $200 buy a wireless LAN card, plug it in to your home PC and switch on. I go home, plug in my machine, switch on and I send out a signal that says find the fiber. The system starts searching. Of its own volition, it searches out the store, the school or the gas station by hopping from one home PC to another. The head end sends back a signal, the best routing is decided, and the extraneous parts are closed down. For $200, one?time payment, I've got 11Mbit/s access to the Internet for all time providing I keep buying my food from that store. Cool, isn't it? The phone companies don't like this idea at all. It gets better. Two companies have reverse engineered a mobile phone chip set to create a cell-phone base station on a PCMCIA card. So imagine what happens now. We get one of these cards and we plug it in to everybody's home, so when I make a phone call, I don't go 5km away to the base station of the mobile phone company, I just go 3m to my laptop computer into an IP network. The mobile phone companies don't like that one. This turns the whole network upside down. And then of course we should watch the kids. In my area kids throw CAP 5 wiring over the back yard, to link their homes so they can play computer games. Kids are now linking computers wherever they happen to be, wirelessly, optically, or by wire. It's going to create a different model and a different world. The cable companies and the telephone companies have one last chance. If they subsume the ISPs into their buildings and the mobile operators and the wireless LAN operators, and put the 802.11 into every street, it will work. If we can go to the store and buy a hi?fi system, a TV, and a PC, take it home and wire it up, how come we're not smart enough to wire up a phone? It doesn't make any kind of sense. The good news for these guys is they can then mine the copper and lead in their ducts and sell it and they get somebody else to put fiber into their ducts. They can lease out the ducts and focus on the business they're really good at, bit shifting. Where does the wireless bandwidth come for all of this? There's a huge chunk of bandwidth that's not being used above 40GHz. Manufacturers are now developing little radar sensors for the back and the front of our cars to help us park and avoid collisions in fog. That technology is good enough to allow us to do short range WLAN communication at 90Ghz. When you get up to those frequencies, it's difficult to get more than about 500m, which is exactly what we want. We want to be able to reuse frequencies. We don't want to interfere with others. Here physics is on our side, technology is on our side and so is the commercial environment, because this stuff isn't expensive. Everything I've said is on my personal homepage. I get about a thousand visitors a day; they take lots of stuff. I have a very simple philosophy to do with networking. If I give this presentation to a hundred people every day like this and you take the ideas and I keep doing it, I can't get to the whole human race in a lifetime. If you will please steal the material on my homepage and repeat it, then in a mere five clicks I will have reached the entire planet. THANK YOU |