Monday, May 15, 2006
Interview with Mark Thompson, Solid State Networks
Our interview this morning is with Mark Thompson, founder, CTO, and chief architect of Solid State Networks (www.solidstatenetworks.com), a Tempe, Arizona-based company that is commercializing BitTorrent technology for use in movie and video downloads. We caught up with Mark to get a better idea of what the company is up to, and how it fits into the content delivery market.
Techrockies: Tell us a little bit about Solid State Networks and how you fit into the market?
Mark Thompson: We have quite a bit of experience from the content delivery of this space. Specifically, we worked quite a bit with Limelight Networks and developed the download manager they use. In the process, about a year/nine months ago we were looking at and watching some of the things that happened with the Half Life 2 launch and thought there was an opportunity address some of the fundamental problems that content delivery networks have from the standpoint of cost and scale. We started looking at peer to peer and seeing what was available and out there. Originally, our intent was to roll our own, and look at what happening and take the best of what was there and start working off of that track. In the process of doing that, BitTorrent was the dominant player. As I started to work more and more with the protocol, I thought it was better to just address the problems that BitTorrent has than actually rolling our own protocol.
Techrockies: So essentially you've put together your own re-implementation of BitTorrent?
MT: Yeah, from the ground up we wrote our own client, servers, and have the lightest weight portable implementations. We've ported to the PC, Mac, Unix, Xbox, the Nintendo DS Handheld, CEll phones--we've literally ported it to everything we could get our hands, to show the flexibility of the actual library. We typically use less than 1 Megabyte of memory per transfer, and less than 1% CPU load for more than 100 peers. That's on a typical class PC. It's very lightweight and fast. The overhead of the library is only 200K.
Techrockies: Describe how you're going to market with this technology?
MT: We get to market in a couple of ways. We have our own skinned client that is a really cool application. The biggest problems we saw with BitTorrent were reliable delivery. With BitTorrent, when you click transfer start, there's no guarantee you'll get anything. If you're using this for commercial reasons, it's pretty key that users actually get what you are sending them. The second was actually quality of service, making sure than a user's experience downloading was consistent. The challenge with BitTorrent is that it goes very fast, but it starts very slow. As more and more peers kick in the faster and faster it runs, and it then outperforms what a HTTP transfer would. So what we do is we use a reliable source with a CDN. We basically work in conjunction with a CDN. We download first from the CDN and benchmark performance. As peer to peer traffic increases, we decrease the amount of data from the reliable source. That guarantees users a quality of service. Third, and most critical, is reporting and analytics, which is important for commercial deployment. In a commercial deployment, if you make a video available, you want to know how many times it has been download, and get the same vision into BitTorrent as you would putting it up on your web server. We allow and enable a vision into BitTorrent, where we can track how many times a video was downloaded. It's very counter to mainstream BitTorrent, which is all about not tracking downloads. Part of that is also most people don't have BitTorrent clients. The typical user experience is daunting, because you have to install things. The more robust clients like Azureus have tons of features, and it's a bit overwhelming. We have a skilled client, with a simple interface--you download a client for a specific transfer, for that movie. It's completely skinned with whatever the movie graphics are and totally custom buttons and is exactly the same for a Mac and PC, so you have the same experience cross platform. So what users will do is download and run the video, movie, or trailer or whatever it happens to be, and when the video closes it deletes itself. That is one deployment of how a user or customer can use it. The second would be is if they already have something on their machine. For example, iTunes. iTunes can integrate our core technology into their client then any downloads would be peer assisted using our stuff.
Techrockies: Why not just use the content delivery networks and their capacity?
MT: No matter how large a CDN is, basically, it's never big enough. There's always going to be a mass of people that are going to come that would be too much for it. I think Crigley recently wrote an article on Desperate Housewives and putting it on the Internet, and how much it would cost and how much bandwidth it would consume, if ten million viewers downloaded it over the course three days. It turns out to be 65 Gigabits per traffic per second, which is an enormous amount bandwidth. And it's probably more than that, since that effectively considers that every one downloaded and started watching at the same time. Typically, there's a wave in bandwidth, typically a lot during the day and less at night. A large CDN can more than handle something of that nature, but what happens if they have two customers with typical downloads--Desperate Housewives is not the only large TV show. That's when you start running into problems. With BitTorrent, and our strategy, peer assisted delivery what happens is the more users get the content, the more that's available on the network, and the less you need to use that network. What happens is the centralized part, the CDN gives you the reliability you need, and the peer assisted part gives you massive cost reduction and scalability that you can support these enormous trends. Typically, we say that what you can expect is a reduction in bandwidth costs of 50 percent or more. Since we guarantee quality of service, you ultimately still use your CDN or whatever your reliable source is, but you can even get more than that if you're willing to diminish the user experience But essentially having the same performance as a CDN or better, it's going to cost you substantially less.
Techrockies: So do you see yourself competing with BitTorrent?
MT: Not really, what I've really seen them doing is taking an aggregator approach, like an iTunes. Azureus is also taking a similar approach. What they are looking to do is leverage the installed user base they already have, and saying we've got X million users, we put this content through it, they're already comfortable with the protocol, this is another way to sell your DVDs. It's the same as how Apple puts your movie or TV shows on iTunes and sell it for a dollar. Our kind of belief is that the aggregators won't be as interested, a place like ABC may try using aggregators at first, but will ultimately deploy their own network. That's where we come into play. We're purely a provider of the technology. We aren't in front of the user. Our motto is it's your customer, your brand, your network. It's not something where you're participating in something larger than what your company does.
Techrockies: Let's talk a bit about your funding. How are you funded, and what stage are you at?
MT: We're self funded--we basically have no immediate needs for capital, though we are looking on the more strategic side of things. We've already had a lot of inbound queries, and we have three launch customers that are also going to be customers going forward. They represent very different areas, which we did deliberately, so we can understand each market before bringing up more customers. So from an investment perspective, what we are looking at is the network aspect a VC can bring, rather than the capital side, though. Capital obviously is nice too, but it's more about the strategic help.
Techrockies: So it sounds like your product is baked and you are going to be deploying this soon?
MT: Yes, the first trial deployments will be in the next week or two. Our first new customer that will be in the next few weeks. They are currently working with the skinned client and doing things they need to do there. The skinned client is actually very cool in that it does not require developer time to make new versions. We spent a lot of time really refining the interface so that you can give it to third parties, and have them do mockups, and do everything to build this client experience on their own. We'll obviously help and we art people in house, but a lot of times these content creators have a lot of art talent in house, so it's easier for them to just do it.
Techrockies: Thanks for the interview!