Choose storage to match application operations
Start from requirements
A technology name alone does not say whether it fits a system. You need to know which data is read together, which updates must be atomic and how much work is expected. Here 60% of operations require related table queries and joint updates of multiple records; 40% use independent documents.
The contract matters more than peak speed
The educational relational adapter supports the full required contract and 700 operations/s. The document adapter supports 1,000, but only this task's independent operations. High throughput does not make unsupported operations correct. These are properties of two specific adapters, not a universal SQL versus NoSQL comparison.
Why cheaper is not always enough
At 600 operations/s, the document adapter can correctly serve only 240 independent operations, even with spare capacity. The relational adapter handles the required 600 but costs more. Selecting one in the palette replaces the previous choice so you compare alternatives for the same application.
What batching changes
A batch groups operations before responding. Here we study only the extra waiting latency: a larger batch cannot fix a missing required contract or create imaginary throughput. Start with suitable storage, then check the latency limit. Both adapters' prices and properties are stated in the model. The stand does not model a throughput gain from batching; this setting only illustrates added waiting time.
Task
600 operations/s. 60% require related table queries and atomic updates to multiple records; 40% involve independent documents. Choose storage and batch size for this contract.
The relational adapter supports the full contract at 700 operations/s for 60 units. This lab's document adapter provides 1,000 operations/s for 30 units, but supports only the independent 40%. These are specific educational adapters; not all document databases lack transactions. Selecting storage replaces the previous choice; two at once are not allowed.
- Read the requirements and choose storage from the palette.
- Compare supported operations with overall throughput.
- Choose a batch size and check every constraint.