問題文
An application needs sub-millisecond reads of a small set of frequently used values, and the source of truth stays in a relational database. Which design fits?
選択肢
- Keep the relational database as the source of truth and put the frequently used values in a memory cache cluster in front of it.
- Add read replicas of the relational database and route all reads to them, because a relational replica reaches the same response time as an in-memory store once the working set fits in the buffer cache.
- Increase the compute of the relational database until reads are fast enough.
- Move the source of truth into the memory cache cluster and keep the relational database as a backup that is refreshed on a nightly schedule.