Facebook: Server management systems 'a big headache'
Open Compute Project goal: stripped and minimalist
SaaS data loss: The problem you didn’t know you had
Collaboration Summit 2012 Amir Michael, manager of systems engineering at Facebook and a key player in the company's Open Compute Project (OCP), used his opening keynote at the sixth annual Linux Foundation Collaboration Summit to decry the current state of server-management code.
Michael said that too many server manufacturers – and the hardware vendors that support them – rely on competing and inefficient server-management technologies to differentiate themselves from their competition. These systems make efficient datacenter design and control a pain in the neck, and provide no real benefit to the end user.
"All these different vendors have different methods to control servers, and they sell on different management features," he said. "It's a big headache, so we're trying to create minimum base of management functions that's more efficient."

If it's not simple, Facebook doesn't want it
This is one of the key areas for the OCP going forward, he said, and one of six areas contributors are working on initially. Other areas include the Openrack project, which seeks to develop a common, highly efficient power source; a storage project to develop multiple connected hard drives; a virtual I/O system that doesn’t rely on Ethernet; and new efficient datacenter and motherboard designs in which every nonessential component has been stripped out.
Michael appealed to the Linux community to get involved with OCP, promising no membership fees and totally open disclosure and sharing of information. The whole project is being funded by selling booth space at the twice-yearly Open Compute Summits, and he promised that no matter how big its booth space, no one company would be allowed to push its own technology onto the OCP standards process.
Above all, he argued, companies need to move away from considering datacenter design as something that is proprietary. By switching to OCP systems, Facebook was saving over a third of the power used by a standard data center, and its minimalist server design saved nearly a quarter of the hardware costs of traditional server racks by stripping out all non-essential hardware.
It's now almost a year since Facebook began pushing OCP as the way forward, and the idea has picked up some traction in the industry. Dell and Intel are firmly behind it, and other hardware vendors and data center managers are also showing support, with Digital Reality already publishing lists of which of its facilities are OCP compliant.
Judging from the warm applause from the assembled Linux glitterati, Michael struck a chord with the open source crowd. Facebook is pushing OCP hard to the open source community, since it needs a lot of clean, open management code – although how quickly this will translate into hard code remains to be seen. ®
COMMENTS
Little boxes on the hillside
The way the server manufactures read this is
"Please design your hardware all the same so you can be replaced the the next up and coming Chinese manufacture that builds his parts at 10 cents less then yours."
It's good for the end user to have a standard interface, but the manufactures will be fighting like dogs for pennies all while dreaming they could have profits like Apple.
The perfect solution
A friend and I came up with the ideal solution to this dilemma: get Zynga to create a new game - ServerFarmVille. Players take on the role of DataCenter admins. Instead of growing plants or zombies, they get points for solving trouble tickets such as disk space increases and virtual server moves. What they don't realize is that their in-game actions have real effects on the underlying infrastructure.
I'm skeptical...
I think this may work, and may even beneficial for some people.
But I think such projects could easily collapse the very moment where one of the (bigger) players has worked out an excellent design (at least in their opinion) which then won't make it into the official specs itself.
So you have set something up which you know will work for you, yet it doesn't get implemented. What do you do ?
My guess is that you then leave the whole project again and continue setting up your environment in the way you intended it in the first place.
Sharing knowledge and experience is often a very good idea, but when it comes to this then I get a weird feeling when people are trying to 'regulate' some stuff.
Could be just me though...

IT infrastructure monitoring strategies
Agentless Backup is Not a Myth
Top 10 SIEM implementer’s checklist
Steps to Take Before Choosing a Business Continuity Partner
Enabling efficient data center monitoring