The DynamoDB book compares connections to DynamoDB vs SQL databases as "HTTP-based model" and "TCP-based model" and implies the TCP-based models are slower at startup.
My brother in christ, what do you think the HTTP runs on?
Also annoying: when talking about why SQL databases are bad now, they imply that storage is cheap, but CPUs are slow.
That's a half truth. Modern CPUs are extremely high performance, they can crunch numbers like crazy. The bottleneck is often the IO, the CPU can only pull and push data to disk so quickly. Which would mean optimizing for storage is more valuable than ever because the less data you need to read, the less your CPU is bottlenecked!
The author also dismisses any concerns of lock in without a second thought. Amazon is always great and loves us, why worry about lock in!
I don't hate DynamoDB, it is pretty cool that I can type `new DynamoDB();` in a file and get an infinitely-scalable database with no additional effort. But if we call ourselves engineers, we need to critically analyze the benefits and risks of technologies.
