A banking app needs reads to trail writes by no more than a fixed number of updates or a set time window, while performing better than strong consistency. Which Cosmos DB consistency level matches?
Show answer and explanation
Answer: C
Bounded staleness lets you cap how far reads can trail writes, either by a count of versions or by a time interval, whichever limit is reached first. It keeps ordering guarantees while allowing better latency than strong consistency. Eventual consistency sets no bound, consistent prefix only preserves order, and strong consistency never lags.
Why the other options are wrong
- A. Eventual consistency has no configurable lag bound, so reads can trail writes for an unspecified time. The scenario requires a defined limit on staleness.
- B. Strong consistency guarantees that reads always see the latest committed write, which exceeds the requirement. It also costs more in latency across regions.
- D. Consistent prefix ensures reads never see out-of-order writes, but it does not promise a maximum number of versions or seconds of lag.