
Amazon DocumentDB Cost Estimator: Size Your Workload in Minutes - Part 2
Amazon DocumentDB (with MongoDB compatibility) is a serverless, fully managed native JSON document database that makes it straightforward and cost effective to operate critical document workloads at virtually any scale without managing infrastructure.
In Part 1 , we walked through Basic mode — sizing a single collection and interpreting the results. That works well for quick estimates, but most real-world workloads don't live in a single collection. Production environments typically span dozens or even hundreds of collections across multiple databases, each with different document sizes, index profiles, and traffic patterns. That's why Advanced mode is the recommended option for doing real world capacity planning or migration work.
Advanced mode accepts a CSV with metadata and workload details for all your collections at once. The calculator aggregates everything — working sets, storage, I/O — and produces a single cluster recommendation with full pricing breakdown. Whether you're migrating from MongoDB or building out a new multi-database deployment, this gives you the complete picture in one pass.
There are two paths to get your CSV ready:
Path A: New Deployments or Manual Data Entry
Use this if you're planning a new deployment or don't have a running MongoDB instance to pull stats from.
Step 2: Manually fill in the CSV with your database, collection, and workload information. Each row represents one collection.
Step 3: Upload the completed CSV to the calculator
Path B: MongoDB Migrations — Automated Sizing Tool
Use this if you're migrating from an existing MongoDB cluster. The automated tool pulls collection metadata directly from your running instance — no manual data gathering needed.
Step 2: Clone the Amazon DocumentDB Tools repository:
github.com/awslabs/amazon-documentdb-tools
Step 3: Run the sizing tool against your MongoDB cluster:
python3 sizing-tool/sizing.py --uri "mongodb://username:password@hostname:27017/"
This generates a CSV pre-filled with your collection metadata — document counts, index sizes, average doc sizes, etc.
Step 4: Manually update the generated CSV with your workload information. The tool can't observe your working set and traffic patterns, so you'll need to fill in these fields yourself:
Parameter | Description |
|---|---|
Index_Working_Set (%) | Percentage of indexes that need to be in memory |
Data_Working_Set (%) | Percentage of data that needs to be in memory |
Inserts_Per_Day | Number of insert operations per day |
Updates_Per_Day | Number of update operations per day |
Deletes_Per_Day | Number of delete operations per day |
Reads_Per_Day | Number of read operations per day |
Index_Working_Set (%) & Data_Working_Set (%) — These percentages determine the memory your workload needs on the DocumentDB cluster. Make sure the result aligns with what your current MongoDB instance actually uses for caching data.
Keep in mind — not all mongo instance memory is available for caching. By default, MongoDB allocates memory for caching data as the larger of:50% of (RAM − 1 GB), or
256 MB
Example: Say your index size is 20 GiB and data size is 100 GiB. You set Index_Working_Set to 100% and Data_Working_Set to 10%. That means your workload needs ~30 GiB of memory (20 GiB indexes + 10 GiB data). If you're running an r5.xlarge or M50 MongoDB Atlas cluster tier (32 GiB RAM) today, the memory actually used for caching data is about 15.5 GiB — that's 50% of (32 − 1). So if your workload runs fine on that instance, your real hot data is closer to 15 GiB, not 30. Adjust your percentages accordingly — maybe Index_Working_Set at 50% and Data_Working_Set at 5% better reflects what's actually cached today.
Operations per Day — You can pull these numbers from
mongostat or the MongoDB Atlas console. They don't need to be exact — ballpark figures are fine for a solid estimate.
Step 6: Configure Cluster Options, Migration Source & Pricing Model
Next, configure the cluster options . These choices shape the recommendation as much as the workload data itself:
- MongoDB migration — Enable this if you're migrating from MongoDB. It adjusts index-size estimates.
- Serverless comparison — Adds a serverless pricing card so you can see break-even analysis side by side with provisioned.
- Region — Pick the region where you plan to deploy. Instance, storage, and I/O rates vary by region, so selecting the region where you plan to deploy ensures the cost estimate is accurate.
- Backup retention — select you backup retention period. The minimum is 1 day, which comes at no extra charge.
- Availability — Choose the availability tier that matches your SLA requirements.
Step 7: Generate the Sizing and Pricing Recommendation
Choose Calculate to generate the recommendation



Interpreting the Recommendations
Once you upload your CSV and click Calculate, the results work the same way as Basic mode — memory-based recommendation, operations-based recommendation (if applicable), serverless comparison, and savings plans. The difference is that the calculator now aggregates working sets, I/O, and storage across all your collections into a single cluster recommendation.
Memory-Based Recommendation
This is usually the primary card. It shows the recommended instance type, node count (writer + read replicas), storage, I/O, and a full monthly cost breakdown. Memory sizing matters most because when your working set doesn't fit in RAM, queries hit disk and latency spikes.
Operations-Based Recommendation
This card appears based on your read/write volume. When both cards show up, a banner explains exactly why they differ so you know which dimension is driving the sizing.
When both cards appear, an explanation banner tells you exactly why they differ, so you can understand which dimension is driving the sizing.
Serverless Comparison
The key number here is the break-even utilization percentage. The calculator converts your memory needs into DocumentDB Capacity Units (DCUs) and tells you the utilization level where serverless and provisioned cost the same. Below that threshold, serverless wins; above it, provisioned is cheaper.
Savings Plans Information
Shows 1-year commitment options with the discounted instance cost, total monthly cost, and both dollar and percentage savings versus on-demand pricing.
Disclaimer : This is not an official AWS product and is provided for reference purposes only. All estimates are approximations and should be validated through performance testing.
Conclusion
Sizing a DocumentDB cluster doesn't have to be guesswork. In Part 1 , we covered Basic mode for quick single-collection estimates - understanding what drives the recommendation and how to interpret the results. In Part 2, we walked through Advanced mode - the recommended path for real-world workloads, where the calculator aggregates dozens or hundreds of collections into one cluster recommendation with full pricing breakdown. Start with your actual workload numbers, even rough ones, let the calculator do the math, and iterate from there until you hit the right balance of performance and cost. Give the Amazon DocumentDB Cost Estimator a try and see what it recommends for your workload.
Want to learn about Amazon DocumentDB (with MongoDB compatibility) ? Join the NoSQL Launchpad in Builder Center — that's where AWS NoSQL specialists hang out, share war stories like the ones in this series, and help teams work through exactly these kinds of problems before their big day.
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article