Tuesday, May 15, 2012

Diablo III was doomed to launch problematically

It's become obvious that Diablo III's launch has been what's technically known as a pig's breakfast. The servers required to play the game fell over under the load, and fans are getting hot under the collar at the problems they're encountering. The trouble is, Blizzard couldn't have done anything about this. This was fated to happen as soon as they decided that the game would require connection to their servers.

The problem is the difference between average load and peak load, and the issue of margins. Blizzard had to have enough servers to cope with anticipated average load, but every server added to their cluster after that comes straight out of profit margins. They therefore need to run with the absolute minimum of headroom. Unfortunately, launch day is inevitably going to result in the highest load your servers will ever see, and once you've got a server, you've sunk its cost. With a game as hotly anticipated as Diablo III, launch day is going to be a huge load, far in excess of the headroom on your farm of servers to cope with average load. You're going to get refused connections, or you're going to get cascading failures, and things plain will not work because you won't have enough computing power to cope - and if you had gotten enough computing power to cope, you'd have wiped your profits.

The issue is further complicated by the lack of price balancing. Unlike Blizzard's other you-must-be-online game, World of Warcraft, Diablo III has no recurring subscription cost. You pay your $60 and that's it. With World of Warcraft, continuing players essentially pay for the servers to remain up, to the tune of $15 per month, and that simplifies things enormously. For one thing, you've got metrics; you can anticipate demand much better. Whereas with Diablo III, demand is harder to predict and so you're going to be shaving your margins to try and run the server farm at just above the meltdown point. Not too far above, because that represents a cost without a reliable revenue stream to back it up.

What does this mean in real terms? Diablo III is a cost centre for Blizzard, not a profit centre the way World of Warcraft is. Their real-money Auction House and the cut they take from it will offset that to some extent, but I doubt it will pay for the full cost of keeping the servers up. So enjoy Diablo III while it lasts; there are beancounters who'll gladly kill it off.

Normally, I'd have put this post in the queue, but that's full to bursting already and this is a short, highly topical post.


  1. Frankly if it were me, I'd be using some kind of compute-on-demand like Amazon as my back-end. That way I can spin up more hosts as I need them, and then decommission them as I don't. I suspect you could do all that automatically. Per-hour costs are higher than fixed (sunk) servers, but savings from non-sunk systems which are merely peak requirements would make the exercise worth-while in the long run.

    You could even have a fixed baseline of real servers and just use Amazon for peak usage.

  2. I disagree with the idea of them killing Diablo 3 off. They've spent over a decade making this game, and I think they'll do whatever they have to to offset the price of servers, because flat-out ending the run of a game everyone's been looking forward to for so long would absolutely destroy Blizzard's image.

    I think the auction house will help a lot; it's essentially their equivalent of TF2's hats. If that doesn't cut it, there are other options; incorporating advertisements into menus and loading screens is just one off the top of my head. And they'll undoubtedly be releasing DLCs and expansions.

  3. The trouble is, Blizzard is not simply Blizzard. Not any more. They're a part of Activision, and that can't leave them unaffected.


After some particularly vile spam showed up, I have disabled the ability to comment as a nonny-mouse.