Facebook’s Open Compute Project

dalaohu

Facebook just released Open Compute Project, their now-public datacenter and server design, optimized for situations in which hundreds or thousands of servers are needed such as the biggest websites and web-hosting companies.

The hardware is somewhat similar to what Google revealed in 2009, but with a different approach to battery backups, and the complete release of the formal design documents so that anyone else could, theoretically, copy Facebook’s design and potentially improve it.

Nothing about Facebook’s design is particularly revolutionary to casual industry observers (except the impressive PSU efficiency). The much more interesting question is why they released this. It’s only going to be useful to a very small number of firms for the foreseeable future, and even then, it’s not as if anyone who wants these server or rack designs can just place an order — they’re just designs.

So why release it?

On a large scale like this — not a small open-source project by good-willed individuals — “opening” something is almost always an effort to commoditize it, leveling the playing field as much as possible and marginalizing competitive advantages that others might have had.

It’s usually in a business’ best interests to commoditize its complements. Microsoft commoditized PC hardware because its software needed a home. Companies that contribute heavily to open-source, such as modern-day IBM, commoditize software because they sell consulting and support services. Google commoditizes applications, platforms, and web technologies because it needs places to put its ads and people to see them. (Google also tries to commoditize anything required to get online: web browsers, DNS, and in some cases, even internet connectivity.) Apple commoditizes apps to make iPhones and iPads more attractive (and exclusive).

Nobody “opens” the parts of their business that make them money, maintain barriers to competitive entry, or otherwise provides significant competitive advantages. That’s why Android’s basic infrastructure is “open”, but all of Google’s important applications and services for it aren’t — Google doesn’t care about the platform and doesn’t want it to matter. Google’s effectively asserting that the basic parts of a modern OS — the parts that are open in Android — are all good enough, relatively similar, and no longer competitively meaningful. Nobody’s going to steal marketshare from Google by making a better kernel or windowing API on their competing smartphone platform, regardless of whether they borrowed any of Android’s “open” components or ideas derived from them. But Google’s applications and services are locked down, because those are vulnerable to competition, do provide competitive advantages, and are nowhere near being commoditized.

We can reasonably conclude from the Open Compute Project that Facebook isn’t trying to maintain a top-secret competitive advantage in hardware and datacenter design, and they’re not expecting anyone else to gain a meaningful, exclusive advantage by copying ideas from theirs and keeping the results secret.

But the benefits of commoditization in this area to Facebook are very small. They probably won’t get many meaningful improvement suggestions. Almost nobody is operating at this scale,1 but even if a bunch of other companies adopted their model, it doesn’t help Facebook much.

My best guess is that this is primarily for recruiting engineering talent. There’s no shortage of engineers, but there’s always a shortage of great ones, especially in Silicon Valley. Google has been a talent vacuum for a long time since it’s so appealing for most engineers to work there.2

With this move, I think Facebook is telling the geek world that they’re just as big and serious of a tech company as Google, and if you want to work on large-scale, interesting engineering challenges that affect hundreds of millions of people, you should work at Facebook.

jerryclark

那个准备要花一年时间写facebook的那位童鞋呢？

dalaohu

估计去IKEA买Rack先了