CHIP DESIGNER Nvidia has explained how to build a Raspberry Pi server cluster.
Nvidia high-performance computing systems engineer Adam DeConinck initially showed off his homebrew cluster last week in a blog post, boasting about a device made up of five separate Raspberry Pi servers, each with its own tiny ARM CPU.
The whole thing uses just 16 watts of power when running at full capacity. Nvidia has decided to let the world in on DeConinck's secret, outlining the step by step process by which enthusiasts can build their own.
Because the Raspberry Pi is driven via a microUSB port, the cluster is run through a five-port USB hub in order to distribute power to all five tiny computers.
"Any powered hub should work as long as it can provide at least five volts and 700 milliamperes to each Pi," DeConinck explained in another blog post. "You'll [also] need a basic eight-port Ethernet switch for networking."
DeConinck's rack was custom built, so he advises custom designing it yourself out of Lego.
"Use one of the Pis as a head node, like in a larger HPC cluster, which provides a shared file system, cluster scheduler and other services," DeConinck added. "The other four Pis will act as compute nodes, used only for running computation jobs."
The most recent version of Raspbian Linux can be used for the operating system. An open source tool called Ansible was also used for setting up automation of all the system software.
After ensuring that his tiny cluster is optimized for speed and performance, DeConinck plans to convert it into a Hadoop cluster for big data processing, and use it to experiment with other system administration problems.
Eventually, he also plans to get Tegra into it, turning it into a portable demo system to be shown off at conferences. µ