OnLive debuted its gaming service at the Game Developers Conference in what was apparently a pretty impressive demonstration, to judge from the press and blogosphere buzz. Basically, OnLive will be running games on its server infrastructure, and then streams them live to users over the Internet, thus allowing users to play titles for multiple consoles, as well as games whose normal hardware specs exceed their own PCs, on whatever computers they want.

Forrester’s Josh Bernoff is enthused about both the announcement and the broader implications of “your life in the cloud”. His take is an interesting read, which I’m not entirely sure I agree with in its entirety. However, I do think that the implications of OnLive’s technology is well worth thinking about in the context of hosted desktop virtualization.

In order for OnLive to be able to deliver graphics-intensive, high-resolution, fast-twitch games over long-haul Internet links, they have to have an amazing, very low-latency way to transmit screen images from their central servers to users at the edge. We know it has to be screen images because in their scheme, the end-user’s computer is not responsible for rendering anything. (This kind of display is a hard problem; previous attempts to display games via remote desktop have run into serious performance issues.) From the way this is written about, the trick is that it’s sending video, meaning that it can stream as quickly as live video in general can be streamed. Real-time screen update is theoretically awesome for business uses too, not just for gaming. So I am extremely curious about the underlying technology.

I’m not sure whether I’m really OnLive’s target audience. I own all three modern consoles (Xbox 360, PS3, Wii), and a lot of my games come with peripherals. So my primary interest in this is mostly the ability to truly get games on-demand. But I am enough of a performance hound to own a high-end gaming monitor, gaming keyboard, gaming mouse, etc. for my PC (although ironically, no high-end graphics card), so any compromise in latency might not be my cup of tea. But it is certainly a terribly interesting idea.