3 ms·
There are plenty of workloads where 20% slack in the workload would result in massive (both relative and in absolute terms) additional costs. In these sorts of
by cthalupa 4y ago
There are plenty of workloads where 20% slack in the workload would result in massive (both relative and in absolute terms) additional costs. In these sorts of workloads, it's incredibly common to run any individual server in the fleet very close to maximum capacity. And even if any individual workload isn't, virtualization and containers letting you bin-pack servers makes it even more common.
- Volundr 4y agoSuch as? Sure if you are operating on massive scale or have a remarkably steady load that 20% margin might be feasible, but that's the exception not the rule. Especially for a lone piece of physical hardware that can't scale up/down.
- eropple 4y agoI'd be interested in an example of this, too, because I've been doing this for a while and I can't think of many--particularly in a modern, scalable environment (because Hetzner's APIs aren't great but you absolutely can scale automatically in their environment if you want to).