>> BUILDING A RASPBERRY PI 3 CLUSTER (PART 1)

This has always been on my to-do list; I have always wanted to build a cluster of Raspberry Pi's.

The Raspberry Pi launched a few years ago as a $25 computer to promote the teaching of computer science in schools and developing countries. With so much power packed into the credit card sized computer - a number of people have always wanted to create a super-computer, or cluster of them. With the introduction of the Raspberry Pi 3 it gets more interesting; as the device already has four ARM cores; so clustering them together makes it even more interesting.

There are a tonne of YouTube videos online giving you step by step instructions on doing this dating back to the original Raspberry Pi - definitely worth watching. I eventually merged bits and pieces from different existing projects to build my cluster - giving it a little more nerdy overtone.

RASPBERRY PI - BASIC SETUP
In order to start building a cluster of Raspbery Pi's - first you will need to configure each one with the latest raspbian environment and make sure that it is updated to the most recent builds. A number guides exist on Raspberry Pi website for installing and updating to the latest version.

This is the most time consuming part; you should prepare as many as your design permits.

RASPBERRY PI - CLUSTER DESIGN
I wanted to keep my cluster simple; utilizing whatever hardware I had lying around and minimize what I needed to buy to get the project working. I had a spare 5-port network switch and tonnes of cables - so, initially all I needed was some Raspberry Pi and a USB powered hub for power. I also picked up a 128Gb USB memory stick to use as a shared disk drive across the devices.

The design is quite simple:

The design is to have one of the nodes act as a master; connected to a public network over the in-built WiFi adapter available on the Raspberry Pi 3. The remainder of the nodes will act as slaves; on a private network that is managed by the master node. A 128Gb USB flash drive will also be used as a shared disk drive using NFS (Network File System) on all the nodes.

The end result will be a single unit connected to three other providing a total of sixteen ARM cores running at 1.2Ghz ready to do some heavy processing, we are effectively building a mini super computer that can be used for parallel programming - a very interesting field in computer science.

In subsequent entries; I will detail how to configure the master and slave devices - and eventually write a sample parallel application to demonstrate the cluster working; I promise it will be more than the "Hello World" examples that everyone else has put together.