Thursday, April 16, 2009

High in the Clouds

Anyone who knows me knows that I instinctively want to attack and tear down over-hyped nonsense. In the past this has included such things as Second Life.

More and more I am running into "cloud computing" as the latest meaningless buzzword.

Well its not really meaningless if you think about it. What after all is a cloud?

A cloud is a body of vapor, made up predominantly of empty air. To create a cloud you need a lot of hot air. So "cloud computing" is actually pretty descriptive of what really exists. Mostly vapor and hot air.

The latest guys to jump on the cloud computing buzz-word and try to ride it like a senior on prom-night are "OnLive." They claim to be able to move your game to the cloud and then (to quote their web site) "instantly sending your controller actions upstream and the results back downstream" over the "broadband internet."

Well, to begin with, as Prof Waldo likes to point out, a guy named Einstein proved quite awhile ago that instantaneous communication over distance is impossible. Even giving them a bit of lee-way on the word "instantaneous" this reeks of someone who has never had to do ANY internet game design. Latencies are an unavoidable reality on the internet. The nature of the beast. Unless they have invented time-travel technology, they are in the same boat as everyone else there.

So until they show me this magical latency disappearing act working reliably in the uncontrolled field, I'm going to count that as one part vapor and two parts hot air. And without that, this whole concept fails. If you've ever played a game that experienced "lag", imagine experiencing that all the time, unpredictably, on every keystroke.

Even if they could make this proposition work, one has to ask "why would you want to?" What they propsoe to do is to take the computing that today is being done on computers we game developers and operaters get for free -- the end user's machines -- and move it all to server horsepower we would have to pay for. Ontop of that they then saddle us with vastly increased bandwidth demands we ALSO have to pay for.

The future of online gaming is clearly the OTHER direction. Figuring out how to use MORE of the free massive parallel network we call our users, not less. There is some interesting research being done right now in Universities in Germany, Scottland and Taiwan exploring how to build that network and utilize it in a way that is reliable and that cannot
be influenced by the owners of the machines. I think practical applications of this work are probably still 5 to 10 years away, but that is clearly the right direction.

In the end, there really is only one possible environment where Onlive's technology can make sense, IF they can get it to work at all. Thats in the hands of the cable providers. They have to pay for the boxes they put on top of your TV, and they ship a lot of them, so they want them as thin and cheap as possible. They also own the bandwidth. So driving bandwidth usage is actually good for them. But thats the only market I can see where the economics *might* make sense. Interestingly enough, since they control the pipes, thats also the only market where they could actually control the latencies of communication to some degree. Instantaneous isnt possible but if they were willing to prioritize the game traffic over all other traffic inside of their closed network, and invest some hefty hardware resources to it, they could probably bring it down to an acceptable level.

But right now, OnLive is a heated mass of vapor. Time will tell if they can manage to be anything more.

My next target in "Cloud Computing" will be our practical experiences with EC2. Watch for it!


No comments: