Interview: David Perry’s Gaikai

Interview: David Perry's Gaikai

Once known for Earthworm Jim and then for bringing Korean MMOGs to the west, David Perry could soon be known best for creating Gaikai, a streaming game service that works in your web browser. We caught up with him at Develop to discuss the project, the realities of lag and publisher attitudes to a technology that could revolutionise the way games are bought, sold and played.

It seems you’re always involved in a lot of different projects at once. Is that a problem?
I’m all over the place. I don’t just do this with the game industry, I do this with life itself. If I get on an airplane, whoever sits beside me, I know we have something in common, and it’s my challenge to find out what it is. My theory on this stuff is, you should try everything. That means I’ve got golf stories to tell, I can talk about parachuting, I can talk about snowboarding, skiing, scuba diving. Whatever it is, I’ll talk about it. In the game industry, it’s kind of like that. You can get stuck in programming, and then, if you have to sit with an animator, you don’t have a lot to say to them. So the challenge is, what can you learn about animation? Who can you talk to, what can you do? What can you do to level yourself up in the world of animation? Then you do the same thing in AI, in production, and that’s generally my mantra.

Have companies ever said, “Oh, we’d like to get Dave Perry involved in this project, but he’ll get halfway into it and then he’ll suddenly become interested in videogame consulting, or geology, and he’ll lose focus”?
Absolutely. It’s definitely a huge risk with me: if it’s not really interesting, I’m not going to want to do it. I get really passionate about something that has an interesting hook, or a new idea. That was the idea with Shiny, and today the core is still the same: I hear about something exciting and I have to get involved.

Can we talk a little bit about Gaikai?
Like the name? It’s a little bit of a trick you can use these days: come up with a name that’s not used anywhere in the world, and then you can monitor how you’re doing. You go into Google and type in Gaikai, and you can see how you’re doing.

And how are you doing? What’s the response like?
I think last time I checked we were at 350,000 pages, but without a single dollar spent on marketing. The chatter’s either: “This is the best thing ever,” from someone who can’t afford to buy the latest hardware, and then there’s the other people who say that it’s completely impossible and I’m just nuts. That’s fine, they just haven’t played it yet.

Is it fair to say that the idea is hostage to broadband penetration?
Absolutely. You’re betting on Moore’s law, and betting that the internet is not going away. To me, those are pretty safe bets. Is it the right time to do it? After playing it myself, I became a believer. It’s possible. In fact, it’s surprisingly possible. A thing that happened to me here at the hotel: anytime anyone plays, we collect the data to build up a picture of what’s going on in the web. Everywhere I can possibly log in, I do, to get a sense of what’s going on. I get to my hotel and the connection’s terrible, which is perfect. I log in and start playing Mario Kart. It’s playing, but it’s skipping a little bit. Our server’s in Amsterdam, so I think, it’s probably not that bad considering it’s going to London and then Amsterdam. Then, I check the analytic data, and it turns out the Amsterdam server’s turned off, and I was connecting to Fremont in California. And I’m playing it from this hotel. That’s a game-changer. It’s actually playable using a hotel room connection to California. Of course, there’s the speed of light issue, so it does add lag. But we would never even make you play to Amsterdam. We’d be thinking really local – there’s always a data centre nearby somewhere.

How much lag is a problem?
It’s very simple. Lag is the time-frame between the data doing a round trip between your machine and the server. If I was to put the server in this hotel, the lag time would be sub one millisecond, so there’s actually no feeling whatsoever. If you go to California, it would be about 150. As far as I understand, the Wii controller is around 100, so you can get an idea of what lag is, and I spoke to the Guitar Hero team, and they said their acceptable lag is 55 milliseconds. This is starting to give you a feeling of what’s going on. Now, my ping time from my office in California to our California server is 20 milliseconds, and that’s the entire length of England. For the actual final experience, we have a server in Irvine, California, which is close enough to my home, and I’m getting a ten millisecond ping time. And ten milliseconds is just phenomenal. You cannot feel a single thing. It’s full twitch.

Does that give you a different problem, since you’re going to have to invest heavily in servers early on?
That’s right. You have two business models. One is a peak business model, and the other is a scaling business model. OnLive chose the peak model. The peak model means you build the network – my estimation puts that at about $150 million to build that – so that you can handle a deluge of players. The reason for that is, you don’t know how many that is, so they’d better be able to handle it, and, secondly, if you pay for a subscription, you’d better have servers available when they show up. Because, if you’re paying a subscription, how happy will you be if you log in and there are no servers available? So, to design for peak, you have to have excess hardware. That’s a bit of a problem, because it’s a huge investment up front. It also means that, if peak dies at night-time when people are sleeping, they’ll have a lot of hardware that’s just waiting. They’re paying for all that power all night long.

Our model’s different, which is scaling. Imagine Gaikai started with just one server. That server gives me about 3,500 plays, if you give each person an hour, on five or six instances per server, per month. So Activision comes along and says, “I’d like to buy 5,000 people, please”. They just took up my whole server. Immediately we need more servers, so we’ll buy another. Then Activision wants 10,000, so we buy another two, and that’s how the model works. Every time somebody starts buying, we look at the demand and buy more servers. Here’s the twist with ours: whenever the demand is own – say Activision goes away – there’s always going to be somebody who wants to pay rock bottom: They’ll be ready to come in when Activision leaves, so the servers will sit at 100 percent capacity all the time.

If you’re adding servers and shifting usage around on the fly, how will you make sure these kind of changes don’t impact people playing the games?
Because initially we’re offering our services as a way to get into new games. We’re offering to sell new players. Would you like to buy new players? If that service isn’t available, we don’t offer it to you. Imagine our servers are at capacity, the button just doesn’t show up. It’s our job to scale quickly and efficiently, as we will, and that’s our whole strategy. Assuming that we do that, now we’ve got the ability to serve games anywhere at any time. What if a website is going to do a preview of a game? How about if the preview version is available with the article? That’s the kind of game-changing stuff.

Where does Gaikai sit in amongst traditional retail? Is the idea that it replaces it?
Not at all. We can serve anything – we can serve Photoshop. It’s like YouTube. The goal is that I can get you to play Spore long before you’d ever drive down to the store and buy it. I’m not targeting the person who’d already buy Spore, I’m targeting the person who would never think of playing it, who just browses and thinks, that looks interesting. Click. I just found you a new player you’d never have had in their life.

But would they then go on to buy it?
Well that’s your choice. Imagine you’re the publisher now. You’ve got me playing Spore, and I like it, and then half way through, we say, “Would you like to save your progress? We’ll need your email address.” And now, we know who you are, we’ve got your analytics, the game you played, and how you did in that game. You could offer them the service – they could keep going, or you could offer them the chance to buy the game right now, or you could offer them a coupon to go to retail and buy it. The publisher can offer whatever they want. We’re not ramming our service down the publisher’s throat. We’re already happy, because you paid us to get that player to the point where they decided what to do next.

How do publishers feel about Gaikai so far?
At E3 all the major publishers came to visit, and I really wanted to get a feeling of if they were in or out. Every single publisher was interested, and three of the top publishers offered to fund us during the demo.

What are the big hurdles ahead?
At the moment, we need to find a strategic partner who will add value. We need to find somebody that wants to invest and can actually add something. We already have options for speculative money – people who are just playing around – and now we want to find somebody who will really move the needle.

Retail can’t be delighted about this.
GameStop already hate me. I’ve made so much money from their store, so I can’t be mean to them about that. I can be mean about their used game policy, because they’re pushing the industry to digital distribution perhaps faster than it would have gone, but that’s all. It’s smart what they’re doing with used games, but it’s not smart if you want to keep in retail.