Facebook unveiled a project called “Open Commute Project” some two years back which meant to share designs related to the hardware used in cloud computing such as racks, storage and servers. In this way many companies could benefit by building their own cloud equipment and not rely on the limited options provided by the vendors who supply hardware.
Facebook started using its own project by designing its own data centers and the related hardware. This has reached a stage where either Facebook designs by itself or allows the designing task to others as per their specifications. Of all the Facebook data centers, these custom made equipment comprise half the population. Latest updates on this say that Facebook will be coming up with a data center in Sweden whose size will be 290,000 square feet and which will be consisting of the servers designed by the company itself which means all servers from the Open Commute project.
In addition to the already existing Oregon and North Carolina data centers of Facebook, this one will have around 10,000 servers. Facebook has put its data centers at approximately 11 places. The various factors that come into play are climate, power, source, labor and taxes. Facebook believes in keeping the servers naturally air-conditioned rather than installing air-conditioners. With the help of this project not only has Facebook been able to start making cheaper data centers but they can now avoid the vendors who supply traditional servers.
Google was the first company to achieve this and now Facebook is doing the same by tying up with the original design manufacturers from China and Taiwan rather than going for Dell or HP who are the original equipment manufacturers in the US. With a huge size, Facebook was not able to rely on one particular vendor because one small fault might result in the company losing a huge amount of data.
Recommended for YouWebcast: The Art of Growth Hacking: Gaining Early Traction by Doing Things that Don't Scale
Hence Facebook is using its own data center management tools to handle design and create its own one. Also, when a data center is purchased from vendors it comes with its own different set of GUI, APIs and interfaces. These results in higher cost due to too many unnecessary instruments and material which results in a high level of complexities for the motherboard. This of course doesn’t mean that Facebook is done with Dell or HP. Even though the majority of its cloud devices and tools has been developed by Quanta Computer, a Taiwan based Original Design Manufacturer, the company is still open for multi-gearing, whether an ODM or an OEM.
Following the footsteps of ODM, OEMs like Dell & HP have started to make designs based on Open Commute specifications and Facebook is in talks with HP to test one of its OEM servers for their Sweden based data center. Facebook annually saves 24% on an infrastructure of low cost and around 38% on the operational costs spent on the staff. The reason is that the same kind of workload is handled, whether it is a server from Dell, HP or an Open Commute Server. It is just a matter of how much work you can take out. Also Facebook doesn’t have its servers being virtualized as the software has got all the hardware devices consumed.
This means that Virtualization lowers the performance level without increasing the efficiency. As per latest reports, Facebook has already published the specifications and designs of its own equipment, motherboards and servers.