Whilst researching how to set up my Raspberry Pi cluster I stumbled across some people across the interwebs who clustered together a bunch of Intel NUCs (Next Unit of Computing).
Why cluster Intel NUCs? You get a powerhouse of a cluster. These NUCs can pack some punch, i7s with 64gb ram… cluster 6 of the 8th gen i7s with 32gb ram and you have 24 cores, 48 threads and a hefty 192gb of ram. All for severely less cost and power usage than some fancy dual Xeon setups. Not to mention they run near on silent and consume comparably low energy needs to run them 24/7.
I don’t have a use case enough to fork out and buy a bunch of Nucs to purely play with this type of setup but was VERY curious to learn an amazing use case in the states.
Enter Chick-Fill-A
Chick-Fill-A is a fast food joint in the land of fast food, United States. Chick-Fill-A uses edge computing in their franchise stores, they deploy a cluster of just 3 NUCs, with instructions so even the least tech savvy owner/operator can plug in the hardware.
Once plugged in the cluster automatically builds itself and launches itself into production.
You can read more about Chick-Fill-A’s Edge Computing setup here: https://medium.com/@cfatechblog/bare-metal-k8s-clustering-at-chick-fil-a-scale-7b0607bd3541
The rest of their blog writings are well worth a read too
The Cloud is Not the Answer?
All of a sudden I realised, migrating everything to the cloud simply cannot be my default answer for an application or service.
Sure, Cloud is by far a better option for many most companies and Chick-Fill-A is no exception. But there are many reasons why a hybrid solution is a great answer to many sticky challenges. Latency and security being some initial benefits. Simplicity in utilising basic hardware for tech-challenged franchise owners at Chick-Fill-A another prime real world example. No doubt there are plenty more.
Next Steps
Learn more about Edge Computing
But also develop my thinking around the merits of cloud computing for all scenarios. I will be looking out for more use cases for Edge Computing, hopefully with some real life examples. Some thinking is needed around further benefits of edge computing over and above latency, security and simplicity also growing a deeper understanding of how I might actually architect a solution.
Build my Raspberry Pi Cluster
For now using the RPIs is a super affordable way to mess around with a cluster setup in a lab type environment. Hopefully it will give me hands on learning opportunities for setting up the hardware, messing around with Kubernetes, probably Docker and I can learn more about elasticity and scaling. Should be a fun project.