Calculate Cosmos DB Request Units (RU/s) based on operation types and document sizes.
Last verified: April 2026
Output will appear here...The calculator estimates RU/s requirements by modeling each operation type: point reads (1 RU per 1 KB), writes (5 RU per 1 KB), queries (variable based on document count, size, and index utilization), and stored procedures. It multiplies per-operation RU costs by the expected operations per second, applies a safety margin, and calculates monthly cost under manual provisioning (fixed RU/s x hours), autoscale (peak RU/s per hour with 1.5x premium), or serverless (per-RU consumption) models.
The Azure Cosmos DB RU Calculator helps you estimate the Request Units per second (RU/s) your Cosmos DB workload will consume based on document sizes, operation types, indexing policies, and query complexity. Cosmos DB charges by provisioned or consumed RU/s, making accurate RU estimation critical for both performance and cost management. Enter your read, write, and query patterns to get a recommended RU/s provision and estimated monthly cost across manual and autoscale throughput modes. The calculator also covers serverless Cosmos DB pricing for intermittent workloads, helping you choose the most economical capacity model.
Your team is migrating from MongoDB to Cosmos DB for a global e-commerce catalog with 2 million products. The initial estimate of 10,000 RU/s seems reasonable for the read traffic, but the calculator reveals that the nightly inventory sync — 50,000 writes in 10 minutes — spikes RU needs to 40,000 RU/s. Instead of over-provisioning 24/7, you configure autoscale with a max of 40,000 RU/s. During the day, it scales down to 4,000 RU/s (the 10% floor). Monthly cost drops from $2,920 (fixed 40K) to $580 (autoscale averaging 5K RU/s).
The 1 RU = 1 point read of a 1 KB document baseline is deceptive. A single cross-partition query scanning 100 documents can consume 50+ RUs, and a write to a document with 5 indexes costs 5x the base write RU. Always measure actual RU consumption with the x-ms-request-charge response header before provisioning.
Partition strategy has a bigger impact on RU consumption than almost any other design decision. A hot partition (where one key receives disproportionate traffic) will throttle at 10,000 RU/s even if the container is provisioned for 100,000 RU/s. The calculator can model per-partition limits, but the real fix is choosing a partition key with high cardinality and even distribution.
Serverless Cosmos DB caps at 5,000 RU/s per container and 1 million RU/s per account. If your workload exceeds these limits even briefly, requests will be throttled. For production workloads with predictable traffic, provisioned throughput with autoscale is almost always cheaper and more reliable than serverless above ~$50/month in RU consumption.
A Request Unit is a normalized measure of throughput that abstracts CPU, memory, and IOPS. A point read of a 1 KB document costs 1 RU. Writes, queries, and larger documents consume more RUs. You provision a certain number of RU/s and every operation deducts from that budget.
Manual provisioned throughput gives you a fixed RU/s allocation that you adjust yourself. Autoscale automatically scales between 10% and 100% of a maximum RU/s you set, and you pay for the peak consumed in each hour. Autoscale costs up to 50% more per RU/s at peak but avoids throttling during traffic spikes.
Yes. When multi-region writes are enabled, write RU consumption is multiplied by the number of regions. The calculator lets you specify the number of write regions to reflect this in the cost estimate.
Was this tool helpful?
Disclaimer: This tool runs entirely in your browser. No data is sent to our servers. Always verify outputs before using them in production. AWS, Azure, and GCP are trademarks of their respective owners.