Last week Facebook announced their open compute project.
As it happens, I was happily attending the OpenSource thinktank at the time. Which was a great chance to have fairly unguarded conversations about all sorts of OSS relevant concepts including open source and the cloud.
So it was great to be there when Facebook “open sourced” Opencompute. If you’ve missed their announcement you can read their thoughts about it on Facebook here. You can go straight to the project home page here.
The OpenSource thinktank is conducted under a bit of a code of silence. So I’m forced to repeat comments without attribution here.
But here are a few of the choice comments:
“This is like the strip miners building a shanty town and calling it progress.”
I personally did not get this one at all. I always assumed that Facebook was a pretty good member of various open source communities as well as of course being behind a variety of their own. However, that’s not completely true.
One very, very large cloud provider put it this way:
“Why should we give back to communities or pay for their support? We know the technology better than they do and we definitely cannot give away our competitive advantage which is our core differentiation.”
This provider has extremely formal processes in place to police their developers so that they do not accidentally allow “IP leakage” and so that they do not pick up any open source that is, for example, aGPL licensed (Wikipedia has a good explanation of the aGPL and the efforts to fix the ASP / Cloud loophole in GPL here). Of course, not giving back is an example of the tragedy of the commons.
Ultimately if you do not give back, you do not get. Your short term gain – such as hoping to retain competitive advantage vs. other cloud provider – works against the viability of the very technical foundations upon which you have built your company.
Which brings us to the next comment:
“Maybe they did this to retain their operations team.”
Actually – that one I’ll attribute to me. While many, many, many companies including Nexenta customers such as Hosting.com and KT are proving that it does not take unlimited money and computer scientists to get to internet scale – I think we’d all agree it does take great operators.
I’m confident that the OpenStack project or Cloud.com or Joyent or Abiquio or VMware or MorphLabs or, a personal favorite Cloudscaling, can help you through software and services to build a solution to compete head to head with AWS. And that software, done right, helps level the playing field. In a year the code that the big cloud providers are holding onto for fear of loosing their differentiation will be out of date compared to the code these Nexenta partners and others in the OpenStack community are using.
But…. the world’s best operators will *not* be commodified. If I were Facebook I’d be worried about holding onto the guys that get little credit and yet do a lot of really tough stuff setting up, operating and, yes, fixing systems often in cold, remote data centers.
And all of us like to think we’re doing more than simply funding legal settlements for our founders That’s a somewhat snarky way of saying that today’s technologists expect openness, they expect transparency on the part of their employers. We certainly see this in the tremendous talent that has joined us from the former Sun and elsewhere including legacy storage vendors.
One more interesting comment about the Opencompute project. It is poor journalism from the generally useful Register to refer to Quanta simply as “a Taiwanese operation that now manufactures more notebooks than anyone else on earth.” Congratulations to Quanta for helping Facebook to design their solutions, for being a joint founder of Opencompute, and for winning this business.
It should go along nicely with your approximately 40% market share in the cloud of servers and with your experience building servers, storage and switching for most of the world’s leading makers of enterprise infrastructure.
Quanta is now a $54 billion dollar a year business – almost as large as Dell in revenues and likely growing much more quickly. In short, as their role in helping kick off Opencompute indicates, Quanta is even more important than their position as the world’s leading laptop maker might indicate. They should be much more than a footnote to the story.
In summary – Opencompute, like Facebook itself, is not without controversy. Peel back the onion and you’ll see an interesting mix of business interests at work including Facebook’s desire to apply a more open patina to their brand which is seen in a fairly negative light across much of the open source community; again, I’d argue one reason for Opencompute is that Facebook’s real critical technical resource is the operators who have figured out how to run their software. The software itself will eventually be commoditized, but the people running it at massive scale are rare.