CloudCostCompareCloud cost mistakes small teams keep making
Most surprising bills have a boring cause. The same dozen mistakes account for a large share of them, and each has a simple fix.
1. Resources left running
A test server, a database created for a demo, a cluster from a workshop. Nothing is wrong with them except that nobody remembers them. Fix: tag everything with an owner, and review the list monthly. See tags and labels.
2. Sizing for the peak you imagined
A server sized for ten times your real traffic, because someone wanted to be safe. Fix: start small, watch CPU and memory for two weeks, and resize. The price catalog makes it easy to see the next size down.
3. Forgetting data transfer
Moving data out of the cloud, between regions and sometimes between zones is billed per gigabyte, and a design that moves a lot of data can spend more on moving it than on storing it. Fix: keep chatty services in the same zone and region, and put a cache in front of public downloads. See data transfer costs.
4. Logs without a retention period
Application logs and metrics grow forever by default, and ingesting them costs as well as storing them. Fix: set a retention period, lower the log level in production, and drop noisy lines. See logging and monitoring costs.
5. Unattached disks and old snapshots
Deleting a server often leaves its disk behind, and backups pile up. Fix: list unattached disks and snapshots older than your retention policy every month.
6. Public IP addresses
Clouds have begun to charge for public IPv4 addresses, including idle ones. Fix: release addresses you do not use, and put services behind one load balancer where you can.
7. A load balancer or NAT gateway per environment
Each one bills by the hour and for traffic. Ten small environments can cost more in networking than in servers. Fix: share where it is safe. See load balancer costs and NAT gateway costs.
8. Paying on-demand for steady work
A database that has run flat out for a year, still billed at the highest rate. Fix: commit to the steady part. See reserved instances and savings plans.
9. Committing too early
The opposite: a three-year commitment on a product that changed its architecture in month four. Fix: wait until usage is stable.
10. Trusting a free tier
A free allowance ends, or covers less than you thought. Fix: read the terms and set a date. See free tiers explained.
11. No alerts
Finding out at month end. Fix: set budgets and forecast alerts on day one. See budgets and cost alerts.
12. Not reading the bill
The bill is the one place all of this shows up. Fix: spend thirty minutes a month with it. Follow how to read a cloud bill.
Once you have fixed these, the next step is a routine: the cost optimization checklist.