Question 60
An earlier incident involved this issue: An application proposes storing high-dimensional embeddings in a small PostgreSQL instance without sizing index memory. Which continuing safeguard would address the risk?
Explanation
Correct Answer
Include vector dimensions, index maintenance, and query concurrency in database sizing. This is preventive rather than a one-time repair; it addresses the underlying failure: Vector indexing and similarity search requirements are excluded from the resource estimate.
Incorrect Answers
This would validate network-level database reachability before application credential testing. It is useful when a service on a private VPC cannot reach a Cloud SQL private-IP address although its credentials are correct, but it does not prevent the issue here: vector indexing and similarity search requirements are excluded from the resource estimate.
For the situation in which a globally distributed Spanner application reports slow commit latency when most write clients move far from the leader region, it makes sense to review client geography and write traffic before changing Spanner instance configuration. Here, vector indexing and similarity search requirements are excluded from the resource estimate, so the safeguard misses the recurring fault.
For the situation in which an AlloyDB application suffers repeated buffer cache misses after a product catalog doubles in size, it makes sense to trend data growth and workload locality before changing instance size. Here, vector indexing and similarity search requirements are excluded from the resource estimate, so the safeguard misses the recurring fault.