Capacity Planning

What It Costs When It Fails

Infrastructure that was correctly sized 12 months ago may be undersized today. Traffic grows. Applications become more complex. Data volumes increase. A server that handled peak load comfortably when it was provisioned may be running at 90% capacity now, with no headroom for spikes and no plan for growth. The cost of discovering this during a traffic event is measured in downtime and lost revenue.

Capacity planning is the process of ensuring that your infrastructure has sufficient resources to handle your current workload with appropriate headroom, and that you have a plan for acquiring additional resources before existing capacity becomes a constraint. It is a forward-looking discipline that requires understanding both current utilisation and projected growth.

The most common capacity planning failure is reactive sizing: adding resources after a constraint has caused a problem. This is expensive, disruptive, and entirely avoidable. Infrastructure changes take time to implement, test, and deploy. A capacity review that identifies a constraint 90 days before it becomes critical provides time to plan and execute the right solution. A capacity review that identifies a constraint after it has caused an outage provides only the option of emergency remediation.

The Right Metrics for Capacity Planning

Capacity planning requires tracking the right metrics over time. Peak utilisation is more relevant than average utilisation. The 95th percentile of CPU usage tells you more about capacity requirements than the mean. Storage growth rate tells you more than current storage usage. The goal is to identify the trajectory of each resource and project when it will reach a constraint, then act before that point arrives.