Grid architecture for a million concurrent users?

Most games design their graphics around their servers and what they think their users are going to be able to have for computers. All the graphics in the game are put on your hard drive and used as necessary.

Here we have an unlimited ability to add graphics, avatars, animations, and everything else. All of that has to be served out to users through the network.

There are some major issues with how SL works. They need to be willing to put more things on people's hard drives. The 'grid' needs to be upgraded. And they need to hire a whole lot more people. But for all the problems, it's not fair to compare them to other games. Other games can put 100 people on your screen because all of them have the same 20 or 30 different costume pieces. Here everyone is different.

In order to devise an architecture for the ubiquitous SL of the future we need to start with an understanding of the current architecture.

Here's a rough draft I whipped up using the online diagramming system at gliffy.com.



It's my first gliffy, sorry for the white space at the bottom.

After Google buys them out, sets fire to the Tao of Linden, and threatens to sanction employees who sign off on new releases that break existing features.

I read that in my local newspaper yesterday... Very interesting to say the least! I wonder how long it will be before a similar chip makes it's way to the market. That would certainly improve lag provided they didn't run 80 sims on it...

I'm leery of mentioning the recent demonstration of a commercially available 16 qubit quantum computer, lest I be accused of being someone that still believes in the flying cars depicted in Popular Science for decades.

There've been some pleasing developments in hardware lately that could improve DRIVE (Distributed Real-time Interactive Virtual Environment) performance; the "solid state hard drives" (SSDs) such as the SanDisk or SimpleTech offerings; 3D processors, in which the actual processors are created stacked on top of each other; using photons instead of electrons for onboard communication; and other such developments

Good old fashioned Fiber Optic Cable, in the guise of Fiber Optic To The Desktop, would surely help to alleviate some lag - if only I could get it!

Throwing hardware at a problem may not solve the problem, but it doesn't hurt.

Can anyone provide a link to a real diagram of the exiting SL system architecture?

Sure, here's the structural design of Second Life.




Oh, whoops, sorry, that's the design structure of Wikipedia.

My mistake.

.

Though my serious comment here is that this is what the Second Life architecture may soon start to look like.

This switchover from a proprietary communications protocol to using standard HTTP for object caching means that plain old boring industry standard squid webpage cache servers are going to be able to host SL world content.


Are you familiar with Akamai? They have localized servers in major cities across the country that offload streaming media content onto a nearby source so your streaming video isn't consuming major bandwidth across the entire country. With a switch to HTTP for object distribution soon Akamai may be able to host localized SL object caches as well.


I look forward to the day where my local Novell BorderManager HTTP proxy cache is going to be able to cache SL world assets and allow perhaps 20 people to share a single 3 megabit connection without any lag problems.

.