DynamoDB Capacity Modes: Provisioned vs On-Demand
What you'll learn: How DynamoDB offers two billing models—provisioned capacity that you reserve upfront and on-demand that scales automatically with usage.
The Two Ways to Pay for Performance
When you use DynamoDB, you need to choose how to pay for read and write throughput. Think of it like choosing between a gym membership and paying per visit.
Provisioned capacity mode requires you to specify exactly how many read and write operations per second your table will handle. You set Read Capacity Units (RCUs) and Write Capacity Units (WCUs) in advance. DynamoDB reserves this capacity for you, and you pay for it whether you use it or not—like a monthly gym membership. If traffic exceeds your provisioned capacity, requests get throttled (rejected) until capacity frees up.
On-demand mode automatically scales with your application's traffic. You pay only for the actual reads and writes you perform, with no need to plan capacity. It's like paying each time you visit the gym—more expensive per visit, but you only pay when you actually go.
When to Use Each Mode
Use provisioned when:
- You have predictable, steady traffic patterns
- Cost optimization matters (it's cheaper at high, consistent volumes)
- You can forecast capacity needs
Use on-demand when:
- Traffic is unpredictable or spiky
- You're launching a new application and don't know usage patterns
- You prefer simplicity over managing capacity settings
You can switch between modes once per day, giving you flexibility as your workload evolves.
Key Takeaway: Provisioned capacity offers lower costs for predictable workloads but requires capacity planning, while on-demand automatically scales with variable traffic at a higher per-request cost.