Estimate Azure Cosmos DB RU/s and Monthly Cost
Enter your expected workload and storage profile. This calculator gives a practical estimate for throughput and cost planning.
Assumptions: 730 hours/month, RU costs are estimated at $0.008 per 100 RU/s-hour (provisioned) and $0.012 per 100 RU/s-hour (autoscale max). Storage estimated at $0.25/GB-month. Actual Azure pricing varies by API, region, and features.
What this Cosmos DB calculator helps you do
Capacity planning for Azure Cosmos DB can feel confusing at first because performance is measured in request units (RU/s), not just CPU or memory. This calculator gives you a fast way to estimate three things:
- Required RU/s based on your read/write workload
- Estimated monthly throughput cost
- Estimated monthly storage cost across regions
It is useful for early architecture decisions, project budgeting, and rough comparison between provisioned throughput and autoscale.
How the estimate is calculated
1) RU/s from reads and writes
The model starts with typical Cosmos DB behavior:
- A 1 KB point read is approximated as 1 RU
- A 1 KB write is approximated as 5 RU
- Larger item sizes scale RU roughly linearly
- Consistency level can increase read RU (especially strong consistency)
After that, the calculator adds a small overhead buffer (15%) to account for indexing, query variation, and real-world spikes.
2) Throughput mode
Provisioned throughput is modeled as a fixed RU/s reservation per hour. The estimate rounds up to common increments and enforces a practical minimum floor.
Autoscale estimates a max RU/s with headroom, then uses your average utilization setting to estimate monthly billed throughput.
3) Multi-region replication
When you add regions, both throughput and storage are replicated. In most designs this scales cost roughly by region count, so this calculator multiplies these components accordingly.
When to choose provisioned vs autoscale
Provisioned throughput is usually better when:
- Your workload is predictable and steady
- You want consistent billing month to month
- You already know your baseline RU/s requirement
Autoscale is usually better when:
- Traffic is bursty or seasonal
- You have daily peaks and low overnight usage
- You want simpler scaling with fewer operational changes
Practical tuning tips to lower Cosmos DB cost
- Use point reads whenever possible: They are significantly cheaper than broad queries.
- Choose a strong partition key: Good distribution prevents hot partitions and wasted RU/s.
- Keep documents lean: Smaller items reduce both read and write RU cost.
- Customize indexing: Excluding unused paths can materially reduce write amplification.
- Cache frequent reads: Application-side cache or Redis can cut database reads.
- Review consistency requirements: Strong consistency can be expensive; use it only where truly needed.
Important caveats
This is a planning tool, not an invoice engine. Cosmos DB cost and performance can vary by API type, region, indexing policy, query shape, server-side features, and traffic patterns. Always validate with:
- Azure Pricing Calculator
- Load testing against realistic production-like traffic
- Azure Monitor metrics (normalized RU consumption, throttling, latency)
Bottom line
A quick estimate is often enough to avoid overprovisioning or painful throttling in early stages. Use this cosmos db calculator to establish a baseline, then refine with real telemetry as your application matures.