Akamai: Neverending end to 'world wide wait'

newsmaker In 1998, Tom Leighton partnered graduate student Daniel Lewin to start a company which vision was to end the "world wide wait".

The MIT professor, Lewin and other research collaborators banked their solution on mathematical algorithms and a spirit of teamwork, hard work and resolve to never give up.

The company went through some "very difficult times", particularly when Lewin, who served as co-founder and CTO, was killed onboard American Airlines flight 11 in the Sep. 11 attacks. The dot-com bubble burst proved to be another challenging chapter, forcing the rapidly-growing business to axe headcount.

Akamai has since grown to become a billion-dollar company and one that is gunning to touch US$5 billion in revenue by the close of the decade. Yet the company has remained true to its original vision and is, in a way, "still at the beginning", said Leighton.

During a recent visit to Singapore, Leighton spoke candidly with ZDNet Asia about Akamai's growth, and the company's role in IPv6 evangelism and the Net neutrality debate.

In 1998, you co-founded Akamai with Danny and a few other MIT associates. How did the name Akamai come about?
Leighton: Danny had a friend who worked in PR and so we asked him, "What should we name the company?" He said Hawaiian names are going to be in. So we got a Hawaiian-English dictionary, picked about 20 words and put them on the board. "Akamai" means clever and cool and intelligent--we liked that, it sounded good, so we picked Akamai.

Akamai was founded at a time when the Internet was truly becoming alive. How different is the company today compared to when you first had the vision to start up the business?
Back then, we had no revenue, only a few employees and a big dream. Today we're a billion-dollar company--we service all the top global e-commerce brands, top media brands and nine of the top 10 banks globally. We carry up to a third of the world's Web traffic. So the dream now has a lot of reality there.

We're doing today exactly what we wanted to be doing--the vision has never changed. Our unique approach of putting our servers everywhere, in as many places as we can, and using distributed algorithms to intelligently deliver content and accelerate applications, building a virtual Internet for the benefit of our customers--that vision is the same. But we're still at the beginning in some sense. We've got a billion dollars now, and we're profitable and growing rapidly, but there's a lot more that we're going to be doing against that vision over the next decade.

As chief scientist, you are expected to predict the future of the Internet and work toward solving tomorrow's problems. What do you see are the challenges five to 10 years down the road, in terms of how Internet content is going to be delivered and consumed?
Security is a huge area of concern--the DoS (denial-of-service) attacks are growing rapidly in number, volume and sophistication. In the last couple of months, we've seen probably three dozens of our customers sustain very large-scale attacks. That's more than we've had in the last year.

There's theft of information on the Internet, corruption of information on the Internet. The bad guys have a lot of capability. It's always harder to defend than it is to destroy. So security is a major area of investment for us and we've got a lot of products that have come out recently, or are coming out this year to deal with that.

Another major area is mobile access to the Internet, particularly in Asia. Of course security is also a big deal here--a lot of the attacks are against sites in Asia and come from bots in Asia. Mobile is exploding right now--we just had the first quarter globally where there were more Internet-enabled mobile devices than PCs sold. There are many countries in this region where the dominant access is mobile and mobile technology will leapfrog landline technology.

That creates challenges because it's not just one device that people are buying, it's a multiplicity of devices. Right now we're in a period where every week or couple of weeks there's a new device announced by a major manufacturer. And these devices use different operating systems so if you're watching video,  for example, it's streamed using different protocols. One device won't take another operating system or streaming protocol. This creates a challenge for the content owner--in the old model they would just send their show to the satellite and broadcast it to everybody. It didn't matter what kind of TV you had, you'd get the signal. That's not true on the Internet.

Today, the person with the content has to encode his movie at different bit rates, in different streaming formats. He's got to worry about different devices people are watching it on and if it will work on those devices. We're investing heavily to make those problems go away for the content provider--just give us the movie or the show like you did in the old days, and we'll make it work on all the different environments out there.

Also, enterprise applications and cloud computing are just at their infancy. As your desktop and corporate applications move into a shared facility of some kind--they are farther away, performance problems will be incurred, security issues get raised. So we're making large investments to, again, make those problems go away for large enterprise customers.

Our overall goal in all this is to make the Internet be reliable, secure, perform well, be scalable and be easy to use for business. We build our own virtual Internet that lives on top of the actual one, and we use all our own protocols for routing, communication as well as application layer protocols to make things be more efficient. So it's designed to confront these problems that are already starting to exist and are going to get a lot more challenging over the next decade.

And the security technologies are developed in-house?
We do everything in-house, we do things with partners as well. Take the way DoS attacks work. The typical attacks we're seeing now are ranging from tens of gigabits a second to hundreds of gigabits a second. If you take the traditional IT approach, you put your Web site and applications in one data center, maybe two or three that are spread geographically. You get a tier-one provider and you buy all sorts of security software. There's no way that works in the face of these attacks. The typical attack is usually a thousand to 10,000 times normal traffic volume. No matter how good your filters are or how capable your provider is, when all that traffic lands at your data center it swamps your pipe and your filter doesn't even get to do its job.

The alternative, of course, is that you could spend a fortune--a thousand times the normal cost--but that's hopeless, you can't spend that much money. With the Akamai approach, we intercept the communications to the end-users where the end-users are, because our servers are near the end-users. There's a ton of capacity there. Say, the bot army is from South Korea, well, we'll intercept all that traffic in South Korea so it won't ever get to Singapore.

That is the only way that I know of that you can defend yourself against these large-scale attacks. It's like fighting a war--if you fight the war at home, there's going to be casualties, it's not good. If you fight the war in the other guy's turf, it's a much better thing to do. With the Internet, all the bandwidth is in the last-mile. It's just orders and orders of magnitude more capacity, and that's where we place our servers so that we get scale, we can deliver at high volume and we can intercept attack-traffic right where it starts.

We have a unique capability--we keep the most targeted Web sites up against the most vicious attacks. Part of that is because we've got enormous scale, the other part is we have plenty of smart people working very hard for a long time trying to stay ahead.

One topic that's on many people's minds at the moment is IPv6. What was Akamai's own experience in the revamping of its infrastructure to be IPv6-ready?
There's a variety of roles we play. We strongly support the transition to IPv6. We're participating in the IPv6 Day, we evangelize about it.

What's interesting, I think, is that IPv6 is at the front and center of people's minds in Asia. It is not so in the United States; and in Europe, probably not. IPv6 is very important to the future of the Internet.

We worked with the U.S. government and we drafted the best practices for industry to follow. We're making ourselves IPv6-compliant as a provider. We're also providing a service to our customers so that they don't have to do anything to their Web site and we'll make it look like it's v6-compliant.

The transition is going to take a decade, at least. During that time, you're going to have to deal with both v4 and v6. Today, and for the near future, the v6 Internet is going to perform fairly poorly just because of how it's hooked up to the v4 Internet which is still the dominant mechanism. We're in the position to actually make those performance problems go away so that for our customers and their end-users, whether they are v4 or v6, we can make it work for them.

Which is most challenging about making these performance problems go away for your customers?
Right now, in many territories, it is convincing the customers they need to be paying attention. In the United States, they are just not interested. It's not a fire that's burning down the house so they think they can worry about it next year. As we think about developing the services that customers are going to buy, right now in the U.S. there's no demand. In Europe, there's minimal demand--there's a couple of customers. Asia, I think, will be the first place where we will have real adoption of that capability because here they care a lot about these things and that's smart.

Why is there the disparity in attitudes?
It's possible the crunch comes here first. Asia is growing most rapidly on the Internet...the region's probably got less IPv4 space to start with. I wouldn't be surprised that the proliferation of devices and the growth here is much higher or will be much, much higher than the rest of the world.

In this region we've been hearing quite a bit about the Net neutrality debate in the United States. Was Akamai ever in danger of being shut down by such regulations?
No.

What are the risks for Akamai if the FCC came out to say that all forms of paid prioritization would be prohibited?
I think what the FCC (Federal Communications Commission) is worried about is that the network is unfair in their access and biased toward its own internal properties. That doesn't mean you can't have various levels of service as long as they are available to anyone who wants to pay for them. There was never any risk to Akamai. We worked closely with the FCC to help them understand how things work with the Internet.

What Akamai does is, first, it makes the Internet better for everybody. Everybody benefits when we offload traffic from the backbone, whether you are a customer of ours or not. It's a good side-effect of Akamai being in the ecosystem.

What Akamai does is get better service for our customers who are the Web sites and the Web applications--we're happy to sell that to anybody, there's equality that way. When we work with the networks that are our partners, we help offload their traffic and improve performance to their end-users. It's a very symbiotic relationship. In the Net neutrality debate, we don't even take sides. We're neutral in the discussion because we sort of help both sides. Whatever the government decides to do, we're going to make it work better.

The Net neutrality discussion is slowly moving over to this part of the world as well. What should policymakers consider when working out such frameworks?
You need to consider the interests of both sides. You want to have the citizens be able to get access to content, you don't want to have the content blocked off unfairly. At the same time, the networks have to have some way of generating revenue to pay for the deployment. So they've got to be able to charge for the access, they have a legitimate interest, too. You need to strike a balance.

In the U.S., some of the networks and cable companies were given monopolies by the government. And when you take that, you have a certain obligation now about how you do your pricing. That's sort of the hook that's being used here: Hey, you've got this monopoly, we let you go into the neighborhood, we didn't let anybody else in--you sort of have to give access to let the other content guys in, you can't use it in an unfair way. But, that said, you've got to let the networks monetize it or they're not going to grow. At the end of the day, you want your population to have access to all the content, and it has to work as an ecosystem so that everybody can make money at the same time.

If you could go back in time and change something in Akamai's history, what would it be and why?
It wouldn't have Danny on American Airlines flight 11, that's probably first.

Well, if we could see the future, we could have timed things better. We were trying to grow very fast during the dot-com bubble days and to get as much growth as possible. But if you knew when the bubble was going to burst, you wouldn't have hired the last few hundred employees you had to fire when the bubble burst.

I think we are on the track pretty close to what we wanted to have from the beginning. Our vision's the same. We thought video was going to hit over IP faster than it has--today, video watching over IP is still about 1 or 2 percent--we thought the base would be larger. Basically, things are unfolding in the direction we designed the company for--we're fortunate that it's worked out that way.