Don't Discount The Unconventional

March 2021

Copyright Geert Weggen, Smithsonian Magazine Photo Contest

One of my favorite things to do is to show people what is possible.

Another thing I love to do is to be resourceful and grab whatever's at hand to solve a problem.

As I mentioned in a previous article - ironically also featuring a squirrel in the photo - MacGyver was my hero. I grew up watching the original one, back in the 80s, and was impressed.

MacGyver didn't have the luxury of waiting through an extended acquisition process replete with delays caused by government shutdowns and award protests. For him it was literally life or death, so I'll grant you that his need was a bit more urgent than what we'd encounter in a typical federal office. He couldn't wait for a custom solution to be developed, he needed something right now.

The thing about MacGyver was that he could solve his problems without waiting for that fancy custom solution.

That's because there were other things around him that he could use and combine in a different way to achieve what he needed. You can bet that, if he needed a tiny stand-alone supercomputer with some custom data science functions, he would not try to locate the one company that built exactly what he needed, then wait for them to customize it, and pay an astronomical bill at the end of the process. No, he'd find other more practical options, and do it himself, likely with some everyday household items. Who knows, he'd probably disassemble a toaster and vacuum, apply some duct tape and be coding his way out of the situation in a matter of minutes.

Some of you might require a portable supercomputer that can be used in disconnected environments, with a low enough power requirement that it can be run off a generator.

You may even need to run Kubernetes or Hadoop on it. It may surprise you to learn that it can be (and has been) done using Raspberry Pi. Meet the Pico 20 Raspberry PI4. Totally off-the-shelf product, yet likely considered unconventional by most standards.

https://www.picocluster.com/collections/raspberry-pi4/.

The Pico Cluster website FAQs state that you can run many of the most common software tools on it, such as Hadoop, Apache Spark, Cassandra, ElasticSearch, SOLR, Neo4J, MySQL and Postgres, LAMP stack running on multiple nodes, and most programming languages including Java, Python, JavaScript, Ruby, Scala, Go and many others. But don't just take Pico Cluster's word for it, do a search on 'Raspberry cluster' and you'll see many examples of enthusiasts building their own clusters for various practical uses, including researchers who can't afford conventional high-performance computing capability, or require better reliability.

How large is it? Well, just as a reference, the white cables you see in the picture are ethernet cables. How difficult is it to configure? Well, besides the hundreds of online tutorials, plus directions from the manufacturer, Raspberry Pi was originally created to teach kids around the world about computer science.

I'm not here to market Raspberry Pi, although I am a fan; I use this just as an example. If you get to know the data and computing communities, you'll find that they are incredibly resourceful and likely have developed solutions that are now widely used and very valuable. Yet most of the solutions they have developed have not been marketed or productized in the traditional sense. Also, it is very likely that few outside of these communities are aware that solutions like this exist.

Bottom line: Potentially, the most valuable things you can do for your organization are to: a) have someone research these kinds of solutions for you, and b) take them seriously when they recommend something that you think is impossible.

Cybele Data Advisory can help your organization find more practical solutions like this.