You may recall a post from a few weeks ago where Dan Ackerman and I reported on our experiences with OnLive, the cloud-gaming service that launched on June 17. We both had good things to say about OnLive. We were impressed that it lets you play PC games on computers not normally meant for gaming, Macs included. We were also pleasantly surprised by the connection speed and the in-game responsiveness, two things about which many people remain skeptical.
Though OnLive made a strong first impression, we still have lots of questions about its ability to handle an increased user load, the future of its MicroConsole set-top box, video quality and other topics. And what ever happened with Dragon Age: Origins and Mass Effect 2? Steve Perlman, OnLive's CEO, answers these questions and others in the lengthy interview that follows.
OnLive's home screen

(Credit: CNET)
Rich: What are some of the challenges you've faced since the launch?
Steve: We got hit with a far bigger wave of people wanting to get on than we expected. We're running at our fall projected subscriber numbers, so adjusting was a bit of a challenge. We were hoping to have a gradual ramp because it takes time to get servers deployed, but we managed to catch up. It was a little uneven in some regions of the country. We were able to bring in some people a little sooner than others just because we didn't have servers everywhere. We still have spot shortages here and there. Dell's been a terrific partner, they make most of our servers for us, and we've been able to keep up. OnLive has been up and running 24/7 since our launch date. We haven't had any downtime, and people seem to be happy with it.
Some things like when you bring up your brag clip list, those 10-second gameplay clips people can make and upload to OnLive, we just never expected to have that many so it takes a couple of seconds before you see them on the screen. That doesn't affect your gameplay at all, but it's a user interface that we have a fix rolling out for.
Rich: Most journalists have reported on a smooth OnLive gaming experience, but in our post commenters were still skeptical about OnLive's ability to handle player load, whether it's with current users or as membership scales up. Is that simply a matter of adding more servers?
Steve: Yes, it is. And attaching more broadband connections so it's bringing in more fiber. To give you a simple example, before E3 we had three data centers. Now we have three data center clusters. We ran out of room and literally ran out of floor space in our data centers, but there are other collocation centers nearby that we're in and we just literally emergency-moved racks to those colos and brought in fiber and turned it on, and that was it, more users are on.
It's different than other online systems. Others, as more people get on, sometimes they slow down because of the way they're sharing resources and one person waits for another person to get off.
Rich: Like World of Warcraft, for example?
Steve: Yes, sometimes they have peak load situations where everyone gets slowed down. We designed OnLive so that you have your own private world, essentially, where the route that you get from your home to the server is always optimized, and it's never set up with contention. Contention is when you have several people sharing the same connection, and as you get more and more people piled on there then you get overloaded. We designed OnLive from the get-go so that we did not have contention routing, rather there are private routes for each individual connection.
Rich: Have you received any angry calls from ISPs wanting to know who you guys are, and why you're killing their bandwidth?
Steve: We have relationships with all of the major ISPs. AT&T is an investor, but we have relationships with the Comcasts and the Coxs of the world. One of the things that we did with OnLive which is quite different is that you usually think of Comcast or a Verizon on the end that connects that your home, there's another end to that wire which is where they connect to either the backbone of the Internet or directly to data centers. It's very rare that a service asks them to connect directly to data centers, but we did that to give people optimal routes. Because we do that, the ISPs actually know exactly what bandwidth they're providing to us and how much we're using.
Rich: Are there any ISPs OnLive has not partnered with? What happens to the routing for users on those networks?
Steve: We have peering relationships with most of the major US ISPs, but of course it is impossible to directly connect with every little ISP. When a user connects from a small ISP, we try to route to it from each of our major ISPs, and choose the route that has the lowest latency and best reliability. Given how many routes we have to choose from, we almost always get highly optimal routes.
Rich: How many users does OnLive have right now?
Steve: We haven't announced, and we won't be announcing our subscriber base, but we're far beyond what we had projected. What can I say, we're surprised. A lot of people were skeptical whether this could work, and we figured skepticism equals cautiousness in signing up. It didn't turn out to be the case, perhaps skepticism may be equal to curiosity.
Rich: Can you provide an overview of the OnLive business model?
Steve: We provide the service for free, you can get demos for free, you can try things out, you can go and then rent things, you can buy things. We haven't done it yet, but it's easy to make it so that a Web site can link right into a game. The nice thing about it is that you can have more than one purchase path. Maybe we refer you to Steam or to DirectDrive to go and purchase the download.
We're super happy if people want to use us a demo service, as a rental service, as a purchase service, as a social network, we have folks that are just watching games instead of TV. Some people have said the only way you're going to use OnLive is if you're buying games, and it's just a little bit silly.
Anybody that's a serious gamer plays games in more than one place. If you're a hardcore gamer and you've got a big rig and you want the highest quality graphics than OnLive is not the place where you're going to play your high-end game. Then again, to not have a huge download in order to trial something before you make the purchase decision, why not? Just click OnLive and give it a go. If you like it, terrific. Download the thing from Steam or order a copy on DVD.
Rich: The service is free for now, but we've heard about a couple of different monthly fees after AT&T is done subsidizing the first year of membership.
Steve: It's not going to be that expensive. As a start-up we have to be completely cautious about what we promise in terms of monthly costs. In March, everyone wanted to know the worst-case number, so we said $14.95. We knew it was going to be less than that. And now it turns out that usage patterns are through the roof.
If you take the worst-case scenario, then we'd have to charge $14.95. It assumed that everybody was getting on at the exact same peak hour. The number of servers we have is based on peak load, not the load through the day, but we had no way to know what the peak was going to be. We had some idea from beta, but in beta we had a different catalog of games. We were worried that once we had new games, everybody would be piling on at the same hour of the day. It turns out that isn't the case.
We have a much more spread across peak load, and one of the reasons has been the variety of different gamers. Some games are casual, we have driving games, we have first-person shooters, we have puzzle games, the audience for each of those different games has its peaks at different times during the day, but the same servers of course are used for different games, so that spreads the load over the day. It means our worst fears about worst case peak loads turned out to be unfounded, and we're able to charge a much lower price. [currently said to be $4.95 a month]
Rich: Another concern we've heard from our readers and from others was the fact that they don't own a local version of the game, either on a disc or downloaded. Some people also felt that the three-year minimum for OnLive hosting a game wasn't quite long enough.
Steve: If someone goes and get the three-year PlayPass and that's not enough because they really want to own the physical copy, then go for it. We hope you enjoyed the demo. Or maybe the thing for them is that you've got a weekend to blow, or some kids visiting for the weekend and they want to get a three-day PlayPass. That may be the same person that otherwise would have rented it at Blockbuster. So we're very happy to accommodate different kinds of people.
We have to put a stake in the ground somewhere. We could put five years, we could put two years. It's less of an issue about the licenses evaporating, and more of an issue of whether or not we continue to maintain the operating systems and the graphics cards to run those games. If a game is tied to a particular Nvidia or ATI card, or if it's relying on a particular version of Windows with different drivers, we can't be sure that those will continue to be available as our servers age and need to be replaced. If it's a popular game that can't run on old hardware anymore, the publishers can do an upgrade for the game. Also, servers usually do last longer than three years, so chances are we'll keep running them. But we have a legal obligation to disclose what might happen. I think the probability of us pulling a game in three years is on the order of 0.1 percent. It's also highly unlikely that a game server will evaporate after three years, but we have to allow for that possibility.
We do expect these games to be available indefinitely. But there's nothing I can do to give someone comfort who really wants to own the physical media. If we told them it was around for 10 years, they'd still say we're not so sure the company is around in 10 years.
Rich: What kind of plans do you have for user data tracking, or even in-service advertising?
Steve: We can have games that are ad-supported. We can also have games that are supported as some kind of promotion related to a product.
Say someone has a car that they're selling and they want to promote the new antilock brakes, they might have a driving game that plays instantly. OnLive is delivered through browsers, and it can run as an executable, and it can also be delivered through apps through cell phones or tables, or integrated into consumer electronic devices. We showed it at E3 working on iPads, iPhones, and on Android devices, and of course on the MicroConsole to the TV. You'll see OnLive built into TV's, and you'll see OnLive built into set-top boxes.
As far as having embedded ads in Web sites, a lot of people run OnLive from their desktop, but a lot of people will connect to the service from the OnLive Web page. It should be pretty obvious that it's easy for us to embed a link in a Web page that makes a game run. The reason we haven't done that yet is we've been paddling as fast as we can as far as the basic service.
You will see in time Web pages that people can just click on a link and they instantly play a game, but there are other alternatives. If you finish a 30-minute demo in the OnLive service, for example, we can include a link to download the game, maybe in partnership with Steam or Direct2Drive, or maybe with the publisher itself. Then there will be another option to refer to GameStop or Amazon, or Toys R Us, which would either tell you what store has the thing in stock or has a used copy of it, or you can buy it directly.
Rich: What about collecting user data?
Steve: We anonymize the information we collect. If you give permission to our customer service for tracking bugs, they can associate data with your account. But for our own analytical purposes we have randomized information. We don't know what a particular user is doing, we just know that this aggregation of nameless, unidentified users are doing something. We can tell you how many people played Red Faction: Guerilla, and how many of them got to the third level, for example.
We have a privacy policy on our Web site (http://www.onlive.com/legal/privacypolicy) that maps that whole thing out. We want to be very careful from the outset to make sure we're very respectful of people's privacy. Obviously the information we collect in the aggregate will help determine the number of servers we provide, but I don't think anyone cares about having their information collected in the aggregate.
Rich: You mentioned the tablets and phones earlier, can you talk about your plans for those devices?
Steve: We want to make it available, the priority is PC and Mac, we have to get those stabilized, get Wi-Fi working, and then you'll see the MicroConsole coming out this fall. We'll figure out what the right timing is for the iPad, the iPhone, and the Android. Nobody can promise that Apple is going to permit an application to run, and we haven't submitted it yet. I'm optimistic they will, but if not, it will be available on the Android. It works very well on both the iPad and the iPhone, with the caveat being that these are touch-screen devices, so the games need to have different controls in order to play them. You'll be able to do all of the community things, you can spectate games, you can chat with people online, but when it comes to playing the games, the publishers need to make some modifications to the controls for the touch interface.
Rich: What about building OnLive into TVs and set-top boxes?
Steve: I can't disclose any details about our partnerships or timing of their products. Of course, it's up to them when they'll do their announcements and release their products. But, what I can say is that OnLive does not add any additional cost to a TV that is already designed to connect to the Internet, other than of course, you need a gamepad to work with the TV.
Rich: Why isn't OnLive allowing Wi-Fi right now?
Steve: I literally apologized to people about Wi-Fi. It was a very difficult decision because it does work with OnLive. The way OnLive works is that it adapts to every single connection and it's constantly readapting. If we have the complexities of people's home Internet service, and then we compound that with the interference that's happening with wireless, it's very difficult for us to diagnose what's going on. And our customer service people had not had enough experience talking with users in order to handle those kinds of problems. So we decided to go slow.
It turned out to be a smart move. We have people not signing up because we don't have Wi-Fi, and if they had there's no way we could have met the demand. We would have had to turn people away. So waiting on Wi-Fi ended up putting a damper on growth that actually saved our butts. But what we are doing now is collecting statistics, we're helping diagnose problems in people's homes, and once we wrap our heads around that, we can open up Wi-Fi. By the fall, no worries, you'll be able to connect with Wi-Fi.
...
Bookmarks