 and events in the world of emerging information. Now, it's time to extract the data and turn it into action. Live from the SiliconANGLE Studios in the heart of Silicon Valley, this is Extraction Point with John Furrier. Hi, I'm John Furrier. Welcome to the Extraction Point with John Furrier. We have a special guest here, John Little, who's the famous from MIT who invented Little's Law, which we'll talk about and what it means for computing. And my other guest is David Floyer, co-founder of Wikibon Research. We're going to go deep dive into the math behind big data. So, John, welcome to the program. We appreciate you coming on. So tell us about what is Little's Law? OK, let's start with something simple. Taco Bell, a Dunkin' Donuts, Little's Law has to do with queuing. And we all wait in line, say, at Taco Bell to get a place our order, pick up our order, and leave. OK, so Little's Law deals with arrivals and rate at which people arrive. The average number that are there at any time in queue and the average weight per person in queue. So we have a relationship. It turns out that the average time in queue equals the average arrival rate of customers times the average weight in queue. And so in queuing terminology, L, which stands for line, and lambda, which happens to stand for arrival rate, and W, which is waiting time per person. So it's L equals lambda W. OK, this has become very much used in the world. And the two major areas have been operations management, which I'll come back to in a moment, and computers. Inside computers, there are a lot of queues. OK, and in computer systems, generally. OK, in operations management, you have a big movement a while back in lean manufacturing. OK, lean manufacturing was a set of rules or opportunities, places to look for in your manufacturing system where you might make improvements. OK, but this was guided. And in particular, I know in these written books, a fellow named Mike George, who was very successful at this and built a company and sold it later to Accenture. So what he did was he said, what you want is low cycle time. You want to get the stuff out the door, manufacture it and get it out the door. So that is low average time in queue. And that is equal to the average queue length divided by the transaction rate. And so what he did was essentially look at these potential for improvement. And he found the major potential for improvement is in reducing the queues, the average queues. And so he would scout around and find things in lean manufacturing which would reduce that. And that meant that he decreased what the computer people call latency, but I call average. And in operations people prefer called cycle time. But it's essentially the time between producing the next item or manufacturer, airplane, whatever. And so this was a very successful operation. And it becomes clear because of Little's Law because Little's Law is essentially a mathematical equation. You don't have to go out and check it in the field. It's true. And so it offers very strong guidance to people who wish to improve their systems. But it doesn't improve the system itself. You have to look at the terms of it and pick on the ones that you think you might affect and take it from there. David, you obviously have an interest in storage and operations around some of these new trends going on in the business. And storage has been the hottest area around cloud and this new social web. And so this is an operations equation. And often in the big data world in storage and Hadoop and in the world we live in, the world data scientist comes in. But there's math involved. What's your opinion on the complexities and how Little's Law is vectoring into it? Because you and I were talking about Fusion I.O. for example, some of the things they're doing with low latency. But moving packets around is a queuing theory. There's arrival times, there's transit. This is math, I mean, manufacturing, computing. How do you draw that together? So one of the challenges of computers has been the mismatch between persistent data, safe data, and the memory where all the operations are done in the computer itself. So you're talking about computers working in nanoseconds, memory in microseconds, and then you have the disk where you actually secure that data in milliseconds. That's 1,000 times longer. That's a huge difference. And so you get these queues. You get queues waiting for results to come back from the disk to make sure that it's safe on the disk. They can't do anything until that operation has taken place. What's exciting in the computer world is the introduction of flash memories which can replace this persistent storage, this disk storage, and operate 100, even higher than 100 times faster. And the potential of that is to improve those systems, improve the queues on those systems, reduce the queuing time on those systems. And obviously the trade-off is the cost, their higher cost than the disk drives themselves. But by applying those to the system as a whole, then you should be able to increase, if I'm right on this, Professor Little, if you're reducing that latency, if you're improving the time on that operational side of things, then your potential is to increase the throughput through those systems. So I posted on Facebook today an article around that The Guardian wrote in the UK, and it was about the internet is over. And I want to bring this up because I think what's relevant about John being here today and this conversation is extracting out a little beeping going on here, extracting out the data around what this really means. The article is written around the South by Southwest big tech conference going on where The Guardian writer, a journalist, went down and said, I went down looking for the next big thing. This is a conference that's all about the next big technology trend, and he said he couldn't find it. And his whole article was, essentially, it's upon us that the world of life, the layer of technology, is now integrated into our life, meaning it's already here. We're integrated into a computer lifestyle. So it's interesting about the Little's Law how it's been applied into manufacturing and, say, retail, you mentioned Taco Bell, queuing theories, lines, et cetera, throughput, operations management, that the computer internet is actually being part of our life. This is a new operating cycle. And Jim Law, a friend of mine, wrote, the obvious trend except that gaming is leading technology even more than in the research labs. When cars and leading technology is now even more into our life, when cars and highways were invented, they went from cumbersome to an integrated part of our life with rest stops, gas stations, and repair shops. That is what is happening here with technology, although more personal and ubiquitous. So he's saying social networks are here. So this concept of data, the speed of data, the latency, is a big challenge for the lifestyle of the workers, productivity. So tie that together to the geek terms. So how does that fit into the life of a user and then has that relate to a company like an HP, a Fusion IO, an IBM, an EMC, et cetera? Shall I have a go at that first? Sure. You can comment on it. To me, what's really exciting about this is that previously we designed the industry design computers around these constraints, this big constraint particularly on the disk drive. They had to do their best to get around this. And they invented lots and lots of different ways and techniques of getting around that particular problem. The exciting thing here to me is that when you're looking at computer architectures and you're looking at ways that you can improve it, the biggest single way of improving it at the throughput of it is going to be to reduce the latency. And it's a direct, if I understand your math correctly on this, it's directly related to it. So if you go from a millisecond to 50 microseconds, you've essentially increased the potential throughput. And then we have a new phenomenon with this called the real-time web, which is real-time data analytics. You have mobility. So John, you're out here still talking to some of the big companies that are impacting the social lives of people. OK, well, a couple of elements here. First of all, I'll corroborate this. Say how Little's law fits into what we just heard. He's concerned about transaction rate. OK. In Little's terms, transaction rate equals average Q in the system divided by latency. So if you want to up the transaction rate, you reduce the latency and improve the latency. And so that's Little's law expression of what we've just heard. So I don't want a personal note. I want to ask you an opinion from you. You're out talking to some of these young developers and some of the biggest, most growing companies in Silicon Valley. I won't say their names, but they're in Silicon Valley and Mountain View. Palo Alto, Mountain View. These are the biggest, most, the new franchises in technology, the big new names. And they're impacting hundreds of millions of people with this new technology, these social technologies. And most of the engineers are young. What have you found in talking to some of the younger generation of computer science folks? Well, I found that the guy sitting next to me was from MIT. Did they know who you were? He asked me about it, an electrical engineering professor. Are the kids young? Are they just unconsciously competent? What's your impression, I mean? Oh, they go where the excitement is, and the best people. Are they feeling Little's law like? For a long time, well, after I spent my time on Little's law this morning, this MIT guy went out and he got into a big discussion with another engineer. He said, that gives me some ideas. Do you create a lot of provocative kind of questions for these folks to think about? Well, I had them prove Little's law. Did they prove it? I gave them an easy case. Yes, they did. Cool, cool. Well, technology, the extraction point here is that technology is changing. The platforms are growing layers upon layers, like trucks and cars had highways in there.