Right-sizing
Right-sizing
Silver
A server is billed for its size, not for how busy it is, so a big machine running at 10% is mostly paying for nothing. Right-sizing means measuring what each resource really uses — CPU and memory, at its busiest moments, over days or weeks — and moving it to the smallest size that still carries that peak with room to spare. Within one instance family, a smaller size usually costs less and runs busier for the same load — so the risk is cutting into the headroom. Done well, it is the cheapest saving there is; done from averages, or from CPU alone, it causes the next outage.
- Use when
- Servers whose CPU and memory stay low even at their daily peak · After traffic, code or the instance generation changes · Before buying a commitment discount — commit to the right size, not today's oversize · Resources nobody uses at all — shut them down or delete them
- Avoid when
- Capacity kept on purpose — ahead of a launch, for failover, or as a standby · Load that swings a lot during the day — let autoscaling change the count instead · Without memory data — CPU alone can shrink a server into running out of memory
- Trade-offs
- A smaller size costs less but runs busier — less headroom for spikes · A resize usually means a restart, so it needs a window or redundancy · Fewer, larger servers are simpler; more, smaller ones scale in finer steps and survive one loss better
- Requires
- How cloud billing works
3 objectives · 12 misconceptions
- predict Predict a server's utilization and monthly price one and two sizes down, and whether the smaller size still carries the peak with headroom.
- explain Explain why a resize decision needs the peak (a high percentile over days), memory as well as CPU, and a target utilization below 100%.
- choose Choose between downsizing, shutting down an idle resource, deleting an orphaned one, or leaving a size alone — and name the risk of each.
- Not true: If average CPU is low, the server is safe to shrink.
- Not true: CPU tells you everything you need to pick a server size.
- Not true: The right size is the one that runs at close to 100% CPU.
- Not true: A server with low CPU is always over-provisioned.
- Not true: Right-sizing is done once, when the server is first set up.
- Not true: Changing a server's size happens live, with no restart.
- Not true: A bigger server is always the safer choice, so over-provisioning is just good practice.
- Not true: A cloud right-sizing recommendation shows exactly what you will save.
- Not true: The lowest memory setting is always the cheapest for a serverless function.
- Not true: A smaller size carries the same load at about the same utilization.
- Not true: A server nobody uses should be moved to the smallest size.
- Not true: An unused server should be kept running, in case someone needs it again.