This article is missing a ton of crucial details for a real storage decision. Redis has a particularly wasteful storage approach because it rewrites everything periodically. Postgres and many other systems are able to avoid moving old data as frequently.
What about persistence guarantees? What about replication requirements? What about backup effort? The list goes on and on. "Faster" means different things when you're talking about workloads that require low latency vs workloads that require high throughput.
A network-attached kv skips a lot of the work that an actual database performs. Maybe that's work that you really don't need, but maybe you'll learn after an outage that you wish you had it.
There is no substitute for measuring real workloads on real hardware.
I feel like this article is just trolling database authors, operators, and users, right?
Comment
Thanks for that mansplaining. I had no idea that there are more nuances to this.
Parent comment
This article is missing a ton of crucial details for a real storage decision. Redis has a particularly wasteful storage approach because it rewrites everything periodically. Postgres and many other systems are able to avoid moving old data as frequently. What about persistence guarantees? What about replication requirements? What about backup effort? The list goes on and on. "Faster" means different things when you're talking about workloads that require low latency vs workloads that require high throughput. A network-attached kv skips a lot of the work that an actual database performs. Maybe that's work that you really don't need, but maybe you'll learn after an outage that you wish you had it. There is no substitute for measuring real workloads on real hardware. I feel like this article is just trolling database authors, operators, and users, right?