CS Edwards

Part 2: Building the RPi Cluster

As I mentioned in an earlier post, I worked my way through 'Build a Raspberry Pi Cluster Computer’ .

My Most Used RPi Terminal Commands: Config Access: sudo raspi-config Restart: sudo reboot Shutdown: sudo halt

The setup required at least a half dozen restarts for various updates and configuration changes to be implemented. Things started to get interesting at the backbone and subnet setup test point. I successfully set up the subnet outlined in the article. 

And was able to ping each node from the Node 1 (master node). 

I ran into a point of confusion around setting up the SSH keys. The SSH keys allow the sub nodes (2-4), to communicate with Node1. 

Part 1: Building the RPi Cluster

How it started...

The source paper provides a reasonably detailed table of components necessary for the project. Source Table:

My setup cost $395.74. In my table below I’ve detailed my components and costs. Section one contains the materials listed in the source paper. Section two contains additional materials that I purchased for my personal set up. I also utilized a USB keyboard and mouse, from another RPi kit, along with an HDMI monitor. I did not have to purchase these so they are not listed. 

There are two main parts of this project. 

Part 1: Hardware Build & Configuration

Part 2: Computing Experiments 

Ha’s source paper gives a basic outline of the materials needed to build a four-node cluster using RPi computers. The paper included a list of components and costs and information on the general architecture, operating system, MPI and programming language. 

The source paper did NOT include detailed instructions on how to setup the cluster. I was on my own here and had to look elsewhere for guidance. 

Fortunately, The MagPi Magazine, had exactly what I needed! I used the ‘ Build a Raspberry Pi Cluster Computer’ article  as a guide. Over the course of two days with lots of problems solving, trouble shooting and caffeine, I was able to successfully complete my cluster build.

In the most basic terms - a sequential algorithm (set of instructions) is executed by the computer in order, one at a time. When designing and writing sequential code a programmer must identify the desired outcome of the algo, then break the problem into discrete (stand alone) ordered tasks (steps) that lead to the desired outcome. 

By contrast, a parallel algorithm has areas where multiple tasks can be run concurrently to solve the problem. This requires a shift in mindset that I am still wrapping my head around. In my algo design I need to get used to asking the additional question of ‘what can be done at the same time’, and figure out where it actually make sense to implement a parallel approach. 

I’m working on it...

As an [extramural] undergraduate computer science student, all  material covered in my course work has centered around sequential computing. Looking for a way to expand my knowledge base and get some hands on experience I began to look for paper/ projects that explored parallel computing and incorporated the Raspberry Pi computer.  

I came across a paper by Ngoc Ha at Montana Tech titled “ Raspberry Pi Cluster for Parallel and Distributed Computing “ that seemed both approachable and (some-what) replicable. 

My goal is to work through Ha’s paper to the best of my ability and see if I am able to replicate their results. This is my first time doing a hardware build and learning material related to parallel computing.

It’ll be an adventure! I’m documenting my progress here on this tumblr site.