David Perry: Cloud nine

David Perry: Cloud nine

By Rob Crossley

July 10th 2009 at 7:00AM

Gaikaiâ??s founder uncovers key details of the service, that latency issue, and how it differs from OnLive.

"I guess I’m coming from the world of paying attention," half-jokes David Perry, the creative force behind cloud-gaming solution Gaikai. "So let’s pay attention and see what’s happened so far."

The internet comes along, he says, and completely shakes the whole tree. People suddenly get digitally distributed music from iTunes, movies from Netflix, and games from Steam. The digital business model blows open.

But online distribution didn’t just happen to become a viable, popular and serious threat to physical media, he says. It became popular because companies made digital distribution easy for consumers to understand.

"A great example is the iPhone. It’s not like you couldn’t buy stuff before on phones. What's crucial is how the iPhone made buying things like Apps really, really easy."

In April, Apple revealed it had surpassed the one billion mark in app downloads, less than a year after the service first launched.

"With us all being as busy as we are these days, it’s important that people get content as quickly and conveniently as possible," he says.

"Every time something gets faster, easier and more efficient, people tend to follow.

"But it’s not that the iPhone is completely frictionless. It was a massive step forward, sure, but you still need to go through the download cycle, and in fact with large files on the iPhone you actually need to connect to a nearby network.

"Completely frictionless is getting anything you want from wherever you are. That’s the holy grail."

So, let’s pay attention, says Perry; an industry veteran with numerous accolades to his name but is gunning for one more - a pioneering force behind cloud-computed gaming.

For the first time, Perry reveals key information about Gaikai; how it will be established, the latency issue surrounding it, and how it can be differentiated from OnLive.


The main issue that's flagged up with cloud gaming is latency. How confident should we be that cloud gaming can host games that don't suffer any noticeable lag?
This is the question I most often get because people tell me the latency issue cannot be solved because Skype can’t even maintain a good phone call, or whatever.

The answer to that is that no one has ever tried to solve the issue in the way we are.

We’re not trying to connect someone from England to California and try to maintain a good stream.

What we’re doing is putting a local server in every major city around the world. We’re going to put servers in those cities based on demographics and demand. Moving ahead, we’ll look for areas that might get latency problems and we’ll put another server in there.

The big difference between Gaikai and YouTube is that it’s not in [Youtube owning company] Google’s interest to put a copy of that video in every major city, which is why we sometimes have to wait for the movie to buffer.

But. imagine the possibilities if Google placed the data of its YouTube movies just twenty miles from your house; it would change everything.

That’s what Gaikai is; it’s a completely different model of server infrastructure.

What we do is look at how long the delay is during play. When someone uses Gaikai, we can instantly see what the connection speeds are - and based upon that data we select the best server for them.

Basically, a latency problem is really an issue of how many servers we’re willing to buy.

OnLive announced it is going to launch five data centres [across North America]. I think that’s too few. We’re going for a completely different strategy, which is to use a lot of data centres all over the country.


How many?
Our launch plans are for twelve, but I am hoping to get a much higher number than that. I want to build a new data center each month on an ongoing basis.


You mentioned phone conversations over Skype. In that instance, a patchy conversation is still tolerable; occasional dips are still tolerable. With controlling videogames, it’s likely that even the smallest, slightest, tiniest delay will not be tolerated, especially if such a delay would result in an unfortunate outcome for players. How crucial is it to completely remove all latency issues?
There’s two levels of latency. There’s one level of it that you don’t notice; found between controllers and TVs where player input is processed.

There’s latency everywhere, and much of it is acceptable. The Wii controller I believe has about 100 milliseconds latency to it.

So there’s a certain amount of latency that’s acceptable. I was talking to the Guitar Hero guys and they told me that their limit was 55 milliseconds, and that’s actually an incredibly short amount of time.

With the whole latency issue, I wanted to get a real world answer to it, so we installed a server about 25 miles from my house in a nearby city. The latency on that server is ten milliseconds. Five there, five back.

Now, the demo I put up on the web saw the connection take a round trip of about 800 miles. It basically works fine over that distance, as you could see in the video I wasn’t playing very well but there was absolutely no latency problems.

Even though there was no real problems over that distance, it’s not what we’re looking for. We’re trying to make people have access to Gaikai over the fastest possible speeds.


How else is Gaikai trying to distinguish itself from its competition?
A pretty obvious difference between ourselves and OnLive is the fact that they’ll never have an Nintendo, Sony or Microsoft game on their system.

OnLive is trying to directly compete with the platform holders. It’s trying to rule the living room. The OnLive team is basically trying to build a portal where you can go and sign up for a subscription and get a piece of hardware to play games on.

The OnLive team are positioning it as something where you won’t need a PS3, Xbox 360 or Wii any more; you can just have their box.

OnLive's model is to try and make a micro-console. If they succeed in doing so, they will take away some market share from the other platform holders.

I have to ask; if they really do succeed and take some market share away from Sony, Microsoft and Nintendo, what will have really changed for the videogame industry? Will anything change?

The answer is no. They’ve moved money around, but it’s not like the industry will see 100 million new consumers, but just the same ones who have moved to a fourth console. That’s why publishers really aren’t excited by the OnLive model.

What publishers are really looking for is ways in which games can get to new audiences.

The scary thing for the OnLive team, though, is that they’ll have to advertise as much as they can to get players subscribing to the system.

That makes the cost of acquisition of players really high, because - by the way - they’ll be simultaneously competing with Sony, Microsoft and Nintendo’s marketing teams. I mean, trade show booth costs themselves are significant.

So, if OnLive finds all this money to properly advertise their micro-console and somebody shows up to the service, if the system’s network is not capable of handling that person, he or she will leave and that will be a very expensive loss. Each person leaving the service will be expensive.

At the end of the day, the OnLive team will have to build a network ready to handle the demand that they’ll be advertising to get. So the network needs to be designed to handle peak demand.

If players experience problems at peak times after subscribing how are they going to feel about that? They’re going to be mad as hell.

If OnLive doesn’t have a network ready for peak traffic, it’s game over.


The video demo of Gaikai itself featured a spot of Mario Kart. Do you expect Nintendo, Sony and Microsoft to support Gaikai?
Absolutely.

What I expect is one of the platform holders will do a trial with us, and the others, when they see what happens, will kind of be forced to follow suit.

If you go to Kongregate's first page of top-played games and count how many unique plays those games have had, I counted them all up to total 61 million gameplays. That’s 61 million plays that did not go to the games industry.

That’s people sitting around in their office not playing on the Wii. So, why not put Mario Kart on Kongregate?

Obviously it’s technically unfeasible on Kongregte, but you could play it on Gaikai. Also, our technology actually loads faster than on Kongregate, because flash games have to download, and then you have to watch adverts.

Mario Kart would get to number one on Kongregate in 24 hours. Tens of millions of plays. New people falling in love with Nintendo at their work desks.


Kongregate is completely free for consumers though. I take it that people would have to pay to play Gaikai?
Well, the way we work is another big differentiator from OnLive. We are a service for publishers to find new audiences. That’s our goal.

It’s a very simple vision, and my goal is to get people playing these games for free, until people either buy the console, buy the game, or decide to start paying for the game.

That’s the biggest differentiator between OnLive and us; we’re not a portal at all, we’re a service.

Let’s look at Mario Kart, for example. If Nintendo chose to charge for people to keep playing Mario Kart, then we supply the service that allows them to do that.

We’re not in any way trying to get between publishers and platform holders. We're not trying to interfere with their policies. We're a service.

A great example of where Gaikai works best is with a game like Eve Online. The team at CCP have to buy their players; they have to convince players that they should download and register and all the rest of it.

If you buy a banner advert on a website like, for instance, Kongregate, and it says ‘click here to play Eve Online’, the team are charged for the click, but all the consumer has done is travel to the official Eve Online page, and then they have to decide whether they should register and download the game.

Let’s imagine that this banner ad is doing great and ten percent of people who go to the Eve Online site end up downloading the game - which by the way would be an extraordinarily high turnover. That means CCP has to buy ten player clicks to get a single download of the game.

As I said, the cost of acquisition of players is very high.

But, with Gaikai, the banner ad would say the same thing - click here to play Eve Online - but clicking on it would immediately open the game. Every single click becomes a play.

People always ask me how many publishers have we signed up. I tell them we don’t have to approach anyone; they come to us.

We don’t have to go to publishers and say [mocking tone] ‘would you please, please license your games to us?’.

I don’t want their games, they put their games on our service. I showed all the main publishers at E3 and let them all play, and three of them offered to fund us during the demo. I can’t really name them; three major ones. Three top ten publishers.


You showed Gaikai using Photoshop on the video demo, what other types of apps do you expect will feature on Gaikai?
Beyond games, there could be all kinds of different things on there.

I’ve actually been contacted by casinos; they want to be able to show games on it, meaning they wouldn't have to send out code around the net; just video, which is great security guard for them.

I’ve been contacted by visual effect companies because they want to allow people to use their tool packages all around the office, and obviously everywhere they go as well for things like presentations.

We’re also getting interest from cyber cafes in China; right now they have to maintain each game on each machine with video cards and all the rest of it. They are thinking about setting up a load of cheap PCs and running their games through Gaikai.

Another thing that we’ve realised is that, because all we’re streaming is essentially video, Gaikai can be played on places like Facebook.


What steps are you taking to ensure that the Gaikai system isn’t replicated?
The steps that we’ve taken are not trivial. We have our own operating system that works really well; on the demo we loaded up things like Photoshop, and the reason it loads so fast is because our operating system is focused on running just the application and not all the other junk that operating systems work through.

Another key idea we have is to run multiple games through a single server; so every time we run, say, two games through our server simultaneously, it halves our costs.

That is actually a big design feature on what we’re doing because we’re virtualising our video cards and audio cards, which is not trivial at all. We did a demo with seven Call of Dutys running simultaneously. The servers and graphics cards were able to handle it; it was insane.