ASAdika SmithCloud app platform

Data platform

Cosmos DB RU/s cost and performance tuning

RU/s tuning is about matching throughput to real demand while protecting workloads that genuinely throttle.

IntermediateRU/sThrottlingAutoscalePartition keyDaily burstRight-sizing

Recommendation

Downsize review

Low peak utilization with no throttling is a good candidate for a measured throughput reduction or autoscale review.

1,440peak consumed RU/s
1,600relative monthly units
decision = utilization + throttles + workload timing + partition behavior

Why it matters

Cost savings and performance fixes point in opposite directions unless you separate low-utilization containers from hot paths.

Field notes

  • A low average can hide a painful daily spike.
  • A throttle count without a workload story is not enough. Pair it with timing and business impact.
  • Rightsizing should produce a specific recommendation, not a vague request to reduce cost.