categories

HOT TOPICS

Subscribe to our Feed

Speeding Up the Internet: Algorithms Guru and Akamai Founder Tom Leighton (Part 2)

Posted on Thursday, Oct 18th 2007

SM: Even though you had entered the 50K competition, and had started talking to potential customers, Akamai was not a company yet?

TL: We really had no desire to make a company out of our research. We were all academics and we liked writing papers, proving theorems, and giving talks. We were not businessmen in any sense of the word. In fact, at the end of the 50K, we were approached by people interested in funding a company and we decided not to incorporate at the time. The decision was based on the fact that we were not sure we had the right business plan.

We had written papers and written algorithms, but we had not written a lot of code yet. There is a difference between proving a theorem and actually coding systems that really work. We were interested in getting the technology commercialized and making the web better, though.

Over the course of the summer, we worked on the business plan some and changed the model to what it is today. We wrote a lot of code and convinced ourselves it could really work. It could be a successful company and be profitable, and we were convinced this was the most effective way for the technology to get out there and make a difference. At the end of the summer of 1998 we incorporated and left MIT, and then sought to get funding.

SM: Tom, two questions now come to mind. First, who would you describe as the “we” at the time? Second, what was the business model you developed?

TL: We got a lot of good advice from Battery Ventures. Todd Dagres was very helpful and he became our lead VC. The old business plan was to sell caching technologies to ISPs. By talking to the ISPs, we quickly discovered they had bigger fires burning. They were going broke and could not afford to pay. Then we started speaking with large websites: CNN, Disney, Yahoo, and discovered that they were very concerned about scalability and performance quality. They had a financial incentive to make their site scalable and to perform well, and they would pay us for solutions. We obviously changed our business model accordingly.

SM: What is the architecture you used to launch Akamai?

TL: It was an object delivery technology. In essence, you could take the objects in a page that are static and cacheable and we would cache those and deliver them locally. It was principally designed in the early days for a static website. As the web has evolved to become dynamic and personalized, rich and driven by applications, so has the technology.

The underlying principle regarding the architecture of the network is the same today as it was then – we want our servers in every city where the end users are. You want the end users to interact with an Akamai server a couple of miles away instead of going through Internet peering points or long distances as that is how you make the service really fast and scalable.

That business model is still unique in the industry today. We are in 900 ISPs and we would like to double that number. I am not talking about peering, we are physically in those ISPs. If you are peering to get to the end users you are on the wrong side of the equation. You want to be where the end users are, you do not want to go through congestion points to get there.

That architecture is where we started way back in 1995, and it is still unique where we are as a company today. To make it work in practice we had to develop a lot of technology. We did a bunch of that at MIT, and a lot after we left MIT, in order to make distributed computing on that scale possible. Back then nobody thought it was possible to do. It was an ivory tower kind of concept. Today it is widely accepted. We have servers in 1600 locations, and 750 cities and 900 different networks. Not all of them are very reliable, but the system as a whole is very reliable.

This segment is part 2 in the series : Speeding Up the Internet: Algorithms Guru and Akamai Founder Tom Leighton
1 2 3 4 5 6 7 8 9 10

Hacker News
() Comments

Featured Videos

Comments

[…] robert@paidcontent.org (Robert Andrews) wrote an interesting post today onHere’s a quick excerptOver the course of the summer, we worked on the business plan some and changed the model to what it is today. We wrote a lot of code and convinced ourselves it could really work. It could be a successful company and be profitable, … […]

Modeling » Speeding Up the Internet: Algorithms Guru and Akamai Founder Tom … Thursday, October 18, 2007 at 10:20 AM PT

Anyone who understands peering and backbone networking will tell you that Akamai’s technology is archaic. There is so much bandwidth on the backbone that you don’t need to be in every city anymore. You just need to be in the Tier 1 cities. The amount of time it takes Akamai to route you to their local server is longer than it takes to deliver the traffic from a couple hundred miles away.

Brett Friday, October 19, 2007 at 10:48 PM PT

Brett,

My understanding is that bandwidth is not the issue; physical distance from the server is, at least in the case of very large files.

abe Friday, November 2, 2007 at 12:29 AM PT