Is Second Life in a deadend ?

If you are a regular Second Life user, you may have noticed that they are currently experiencing significant slowdown and shut downs, almost on a daily basis. Access to the service is very often closed for "maintenance" or "emergency actions on the grid". It is said that they have a huge architecture issue and that SL might not eventually be that scalable, with about 3 users per server !!

My guess is that the architecture was designed a long time ago (SL is 7 years old after all) at a time when the company had little funding form VCs (let alone customers), especially after the dotcom crash the 3 years of nuclear winter after that. So they maybe had to cut corners, choose an architecture that was easier to implement in the first place, require less fixed costs. They maybe thought that dealing with huge scalability issues would be a good problem to solve later. Of course all this is pure speculation but this is what I would have done with limited funding.

Now they have to find a solution in an emergency (as described here). Problem is that this kind of architecture issues cannot be solved quickly, even with lots of money.

Maybe it is time for someone else to come up with a better more scalable technical solution ? Maybe Google with Google Earth, who knows ? Or Sony ?

Anyway let's hope that Linden Lab finds a solution, SL is great and they deserve to survive and thrive. (by the way you can find me in SL under "Mathieu Noonan")

Comments

[Mathieu Nouzareth]

I was not saying that 3d scalability cannot be achieved, I am just *guessing* that SL took the wrong choices a few years ago.

I have managed lots of technical projects in my life and I know that a bad choice in the architecture early in the process can sometimes have disastrous consequences later on. Even SL themselves admit that they are currently trying to change the motor of the plane while flying, which is always very difficult to achieve, sometime impossible.

[Van Schroehnberg]

I disagree. Scalability to update 3D graphics is a very viable solution assuming you built your engine with scalability in mind. Take a look at Everquest, America's MMORPG father figure. Its graphics were updated approximately 4 years after it was released. Sony realized that the graphics were outdated and they were able to upgrade with minor problems.

The biggest problem with Second Life is the expanding playerbase and data influx. The world is constantly growing because users are able to constantly build virtual goods which takes a toll on your system.

Hardwarewise, Second Life's graphics are not complex or up to date to today's standards, but the exponential growth of the virtual world definitely takes a toll on the user's computer.

With that in mind, the developers are limited to upgrading the graphics of the platform not only because of the cost, but because it would severely limit their player database due to upping the computer requirement specs beyond the norms.

[Louis van Proosdij]

Well, I'm not sure SCOL can be "revived" 5 years later, but there is lot of inspiration to take from SCOL, and huge experience from key people who worked on the engine and optimizations, to quickly setup the foundations of a new solution.

[Mathieu Nouzareth]

I think it is too bad that Deuxieme Monde did not survived. You were right too early (which means unfortunately being wrong...). Maybe someone should revive SCOL ?

[Louis van Proosdij]

Scalability in 3D virtual worlds is something very complex, and a problem that need to be addressed at the core of the technology. The later your address it, the harder (and highly costly) it is. That's something we experienced at Cryo-Networks with our SCOL technology (1998-2002), based on our initial experience with "Le Deuxieme Monde" (The Second World) for French company "Canal+" (started in 1995, launched in 1997).
We did a sound engineering work addressing the load balancing and optimizing the technology to support thousands of players per server (2002 servers, not 2007 Xeons), and we succeeded. We developed a huge MMO RTS game "Dune Generations" based on Frank Herbert Dune licence, and were handling 2000 players/server during alpha tests.
Alas Cryo-Networks did not survived from the financial crash of parent company Cryo-Interactive. SCOL is now open source, but a bit outdated.

Why did I tell this story? Uh... yes, was to agree with your analysis Mathieu.