Provisioned Throughput Considerations - Comparing the Use of Amazon DynamoDB and Apache HBase for NoSQL

Provisioned Throughput Considerations

Factors that must be taken into consideration when determining the appropriate throughput requirements for an application are, item size, expected read and write rates, consistency, and secondary indexes, as discussed in the Throughput Model section of this whitepaper.

If an application performs more reads per second or writes per second than a table’s provisioned throughput capacity allows, requests above the provisioned capacity will be throttled. For instance, if a table’s write capacity is 1,000 units and an application can perform 1,500 writes per second for the maximum data item size, Amazon DynamoDB will allow only 1,000 writes per second to go through, and the extra requests will be throttled.

Tip

For applications where capacity requirement increases or decreases gradually and the traffic stays at the elevated or depressed level for at least several minutes, manage read and write throughput capacity automatically using auto scaling feature. With any changes in traffic pattern, DynamoDB will scale the provisioned capacity up or down within a specified range to match the desired capacity utilization you enter for a table or a global secondary index.