The on-demand model supplies computing resources or services when requested, without advance provisioning or long-term commitment. In AWS, a virtual machine can be requested and ready within minutes, and temporary test environments can be created only when needed. section 1.3.6
DynamoDB offers on-demand capacity mode alongside provisioned capacity. On-demand mode adjusts read and write capacity automatically; provisioned mode requires capacity configuration in advance, with optional autoscaling. section 12.11
Choose between convenience and predictable capacity costs:
- On-demand capacity reduces maintenance and suits workloads with substantial traffic variation or spikes. section 12.11
- Provisioned capacity can be cheaper when configured capacity runs near full utilization continuously; an idle provisioned baseline can waste money. section 12.11
- A reported DynamoDB migration from provisioned to on-demand capacity reduced monthly costs by 90 percent for the marbot chatbot. section 12.11
Business-hours workload with a batch spike
An application normally handles 100 reads and 10 writes per second from 9 a.m. to 5 p.m., then reaches 1,000 reads and 100 writes per second during a one-hour batch job. Provisioned capacity must cover the peak, while on-demand capacity follows actual usage. section 12.11
Table 12.1 Comparing pricing of DynamoDB on-demand and provisioned mode (view table figure)
Compare monthly DynamoDB costs for fixed read/write throughput under on-demand and provisioned modes. section 12.11
