Innovation at Baidu’s Cloud Computing Data Centers
March 8, 2013
Grazed from DataCenter Dynamics. Author: Editorial Staff.
Baidu, the largest Chinese search engine, is pushing the boundary of data center design with its cloud computing data centers. By integrating new hardware technologies including ARM-based servers, custom-designed all-in-one racks, 10Gb top-of-rack (TOR) switches and self-designed Solid State Disks (SSD), Baidu is boosting the performance of its cloud computing data centers. And it says it expects such efforts will lay the foundations for the technology benchmark for China’s internet data centers (IDC) of the future.
Self-designed 10Gb TOR switch
Baidu deployed a self-designed 10Gb TOR switch on over 5,000 servers in one of its cloud computing data centers. It is a server cluster which is believed to be the largest 10Gb TOR based cluster in the local market. By using self-designed hardware and software, an original design manufactured module, as well as DAC (direct attached cable), the cost of Baidu’s 10Gb TOR is almost the same as the commercial 1Gb TOR switches, Baidu said. Baidu started its research and development on the design of TOR switches in 2011 and launched first generation 10Gb TOR switch in 2012…
ARM on a large scale
One of Baidu’s cloud computing data centers in Nanjing is said to be the first to deploy ARM-based servers on a large scale internationally. By integrating the X86 and ARM platform, adjusting and migrating codes according to instruction sets, and improving debugging tools through the ARM server ecosystem, Baidu says the cloud computing data center gives full play to the advantages of 32-bit ARM servers. It says the power requirement of ARM-based server is much lower than even the most energy efficient X86-based servers. Sources from Baidu say that ARM-based servers enable the data center to reduce total cost of ownership (TCO) by 25% and enhance storage density by 70%…
Read more from the source @ http://www.datacenterdynamics.com/focus/archive/2013/03/innovation-baidu%E2%80%99s-cloud-computing-data-centers&u=2911