Two months ago, my research adviser asked me the question "Can you do it?", I was reminded of the first time he asked me that same question. By the way, I am a researcher, I work on problems using different computational techniques. Back then, we were discussing my research problem and I was trying to beat a deadline for a local conference where I get to present my research. When he threw me that question, I answered him how I will solve it, I started discussing the algorithm. He interrupted me and clarified to me what he really meant, "If you run your program right now, will you get the results before your deadline?". I knew how to solve my problem, but I wasn't sure if I had enough resources. Those people who work with computers, surely know that the computing power means everything. During that time, my research adviser gave me a task to build a Beowulf cluster out of the computers in our laboratory. I didn't know why I was doing that task, not until I was put in a situation where I had limited time.

I have delivered most of my research outputs using my laptop, sometimes, I would divide the computing tasks and run simulations on my designated computer in the lab. For the past weeks, I've been having interest on buying a computer server, or a supercomputer, or anything that would have that grand computing power. But reality hits me that I don't have the financial resources to buy those costly computers. 😅That's when I remembered the cheapest computer I have worked on in my research life.

I listed the notable specifications of this computer, aside from its size, it only consumes a little power to operate. From the projects that used RPis to build their own cluster computer, enthusiasts and hobbyists enjoy the fact its cheap and does not cost much maintenance unlike those huge clusters. You can have your own computer with just $25-$35, depending on the model you want to buy. Isn't it amazing?

Looking at the computational power of this computer, the processor and the 1GB RAM can't possibly do those complex tasks you do on real servers and supercomputers. But the thing is, this palm-sized computer can be clustered with other more RPis, which means we can build a computer that can perform the computing tasks we want to do from the combined computing power of the many RPis we use.

Another advantage of using RPi to build your own cluster computer is that you'll be able to know the process of how to build and how a cluster computer works. You can just shift to upgrading your cluster with better hardware when you finally understand how it really works.

After doing a lot of reading online, I have listed the other necessary things I have to secure like a power hub, wi-fi router/switch, ethernet cable, micro usb cable, sd card, and server rack. I can actually source all those things online but my budget's really tight. An RPi server rack costs as much as one RPi, so I decided to design my own server rack. The photo below is the sketch of what I have in mind, we have a laser cutter machine in the lab so I can build this one using some scrap acrylic sheets. On my next post, I'll show an update of what the finish product looks.

2018 when I was able to build my first cluster, we had to disassemble everything when we moved out of our previous workstation. I wasn't even able to perform a computational task using this cluster, all I was able to do that time is send a Hello from the processors of the three computers I have clustered. This time, as I start from square one again, I hope I could really put it to good use.