I regret building this $3000 Pi AI cluster

Frankly, always thought about Pi Clusters as a nerd indulgence, something to play, not to do serious work.

After a few years of experience with them I agree for the most part. They are great for individual projects and even as individual servers for certain loads, but once you start clustering them you will probably get better results from a purpose built computer in the same price range as multiple pis.

I think the only exception is specifically for studying network/communciation-topologies. I've seen a couple clusters (ca. 10-50 Pi's) in universities for both research and teaching.

It reminds me of the Beowulf clusters of the 90s-2000s, that were all the rage at some point, then slowly lost ground... I remember many friends tinkering with some variant of those, we had one in Uni, and there were even some linux distros dedicated to the concept.

Oh yeah, the "imagine a Beowulf cluster of these" Slashdot meme! I miss those days. At least the "can it run Doom?" meme is still alive and kicking.

Ditto! It reminded me of the time in college when I built a Beowulf cluster from recently-retired Pentium II desktops.

Was it fast? No. But that wasn't the point. I was learning about distributed computing.

There are so many network emulators you can use, such as Mininet or GNS3.

I'm sure pedagogically speaking it's better to use physical devices

Beowulf style clusters went on to dominate supercomputing.