|
ZDNet has an article on Second Life's server architecture... and it discusses its business model and other things.
'Second Life': Don't worry, we can scaleQuick Aside - Linden Lab's Second Life has got to have the most Free Press Per Player/Subscriber of any online game. The article raises some interesting questions about the scalability of the design as the average utilization is about 3 players per server as opposed to 116 players per server (a figure provided by Sony Online Entertainment for Everquest II). Both games use a geographic model for distributing people. Second Life's servers each are dedicated to 16 acres of space while in Everquest II, and most MMOs, there are multiple server clusers that replicate the world - think Disneyland, Disneyland Florida (Disneyworld), etc. that are essentially identical. Second Life - 2579 servers - dual core AMD - total population 240,000 registered users, * substantially fewer paying users Everquest II - ~1100 servers - dual CPU x86, 37 clusters, 20-40 CPUs - total pupulation 250,000+ subscribers Conversely, Three Rings' Puzzle Pirates runs on 4 CPUs (servers?) with about he same number of peak users as Second Life (it is worth noting that Second Life has a very low peak utilization compared to its total registered users, at least when compared to Eve Online a while back). - I can tell you which one I'd rather own stock in. The comparisons are legitimate. First, from a business model point of view, Second Life is closer to Puzzle Pirates than Everquest II. Both are free to pay, pay-for-stuff business models. This free to play model seems to run around a 1-3% conversion rate, though higher payments for those who convert - averaging maybe $30/month for the paying players ... your mileage may vary. Linden Lab claims its infrastructure scales based on "land purchases", the primary business model that the company uses. The business scalability problem with this is that land purchases are far from uniform. Thus, many servers may not have nearly enough users to pay for themselves. Linden compares its scalability to Google or Yahoo, but both of those companies can pretty clearly determine how much revenue per server they are going to generate per month.... multiply those darn ad impressions per server type (mail, search, etc.) by average revenue per ad impression and you've got a business model. For Linden Lab, since it is a land-rent model, the number of renters is some fraction of the game's total active population... Also, I am pretty sure that the rent is uniform even if the sales price is variable, so Linden Lab doesn't benefit when "hot" real estate markets occur (except from the purchase of Linden's currency). The technical scalability problem comes from the uniform geography model that the game uses. Second Life, and most games, have players highly concentrated in some areas and none in others. Thus, "hot" parts of the game can suffer tremendous lag and eventual failure as the game's total population increases. Unless Linden Lab can "split" a server to serve smaller geographic tiles, this doesn't seem sensible. Also, it is not very dynamic - why have CPUs running when no one is there? Traditional MMOs can address this scaling problem by adding additional servers to "hot" clusters/shards and encouraging players to move to different shards (limiting shard mobility is a key load management tool for a traditional MMO). They also are gaining revenue from every subscriber every month, providing a solid base for computing server and bandwidth and personnel costs. Free-to-Play games are a rapidly growing portion of the online gaming market (with Asian games in the lead). These games must be much more sensitive to infrastructure costs than traditional MMOs since they do not have direct, self-sustaining revenue models. Games like Puzzle Pirates and Second Life or, at the other extreme, Guild Wars need to plan and manage their infrastructure cost very carefully as they drive success and profitability. By Daniel Terdiman URL: http://news./2100-1040_22-6080186.html Last March, Cory Ondrejka, the chief technology officer at "Second Life" publisher Linden Lab, bet a symbolic quarter that his virtual world would within two years have more users than the wildly popular online game "World of Warcraft." The bet was certainly ambitious. After all, "WoW," as fans call it, currently has more than 6.5 million users. "Second Life" has 240,000 registered users. But whether Linden Lab's virtual world can catch "WoW" isn't the most pressing question about the virtual world's future for some people familiar with its computer network. "The underlying architecture of the Internet and of 'Second Life' is perfectly scalable." --Linden Lab CEO Philip Rosedale Their concern is more about technology: Can the computer network of "Second Life," using an unusual configuration that dedicates each server to a sliver of virtual real estate, scale with growing demand? "Second Life" currently runs on 2,579 servers that use the dual-core Opteron chip produced by AMD. Each server is responsible for an individual "sim," or 16 acres of virtual "Second Life" land. At peak usage that means that each server is handling about three users. "Most (massively multiplayer online games) have hundreds to thousands of players per server machine," said Michael Sellers, who runs Online Alchemy, a provider of artificial-intelligence tools for online games. "Is there a way they can achieve (significant) elements of scale? I haven't seen that." There's little question that "Second Life" manages far fewer users per server than other virtual worlds. Sony Online Entertainment's "EverQuest II," which has more than 250,000 users, runs on about 1,100 dual-CPU, x86 (x86 is the processor architecture used by most AMD and Intel chips) servers spread across 37 clusters of 20 to 40 servers. Each of those handles around 116 users at peak usage, according to figures provided by SOE. Big bucks needed? These wildly different figures have some observers scratching their heads and wondering if Linden Lab is going to have to spend big to keep the "Second Life" network growing. "My understanding of (Linden Lab's) back-end requirements are that they're absurd and unsustainable," said Daniel James, CEO of Three Rings, publisher of the online game "Puzzle Pirates." "They have (about) as many peak simultaneous players as we do, and we're doing it on four CPUs." But Linden Lab executives have a message for worrywarts: Relax. "It works just like Google, where each (server) is a single, cheap (server) that basically operates and is automatically deployed by our systems and simulates the systems," said Linden Lab CEO Philip Rosedale. Rosedale argued that his company's architecture mirrors that of the Internet itself, which he characterized as millions of servers running in a decentralized system. Linden Lab is constantly adding new servers as its user base grows and as users demand new "land." And since "sims" generate a minimum of $200 in monthly land-use fees, Rosedale contended that the large number of servers pay for themselves. "Can it scale indefinitely? Absolutely," Rosedale said. "It can scale to infinity. The underlying architecture of the Internet and of 'Second Life' is perfectly scalable." He said that most massively multiplayer online games, like "World of Warcraft" and "EverQuest II," are designed around a central database that does the heavy lifting of managing as many concurrent users per server as possible. By comparison, the "Second Life" environment is spread across its many servers, which Rosedale said are in a "tiled network" whose demands on the central database are akin to that of e-mail. "We just throw new machines at it all the time," he said. "So it is we who have the scalable architecture." While there may be questions in the online-games community about Linden Lab's server strategy, the model has proven successful for other companies. "It works pretty well for Google and Yahoo," said Gordon Haff, a senior analyst at Illuminata who was not familiar with Linden Lab's architecture. 'Radically different' approach "It sounds like an approach where they can segment the tasks by segmenting the data structure," added Dan Kusnetzky, formerly the vice president of systems software research for IDC who is now executive vice president of marketing at Open-Xchange. "And that sounds like a good tradeoff." Kusnetzky agreed with Haff that other companies have succeeded with Linden Lab's model. "You can get some unbelievable scalability stories if you can think through the stories and build a lightweight architecture," he said. "That's how Google and Yahoo do it." In any case, some say "Second Life" is already bigger than they ever expected. "They're succeeding because of their radically different approach to this business," said Edward Castronova, an expert on virtual worlds and an associate professor of telecommunications at Indiana University. Indeed, while most online games make money by charging every user a monthly fee, "Second Life" is free to play unless a user wants to own land. Linden Lab makes its money off of land-use fees, the sale of its virtual currency and monthly fees paid by land owners. Rosedale said Linden Lab isn't yet profitable, but soon will be. He also acknowledged that "Second Life" has a difficult user interface that is an impediment to massive adoption and that Linden Lab has to work on that. He pointed to potential future plans to let users create their own "skins" for the interface, a step that would give control over the interface, like all other "Second Life" content, to users. Castronova, who said he does have some worries about the "Second Life" business model, said it's worth sticking around to find out what happens. "Regarding (their) business model, I have the anxiety of someone who went out to explore a river," he said, "and I'm already 200 miles further than I ever thought I would get and there's still more river. Scary, but I have to keep going." |
|
|
