github.com/idealo/mongodb-benchmarking

MongoDB Benchmarking Tool - mongo-bench

mongo-bench is a high-performance benchmarking tool written in Golang designed to test and measure MongoDB insert, update, delete, and upsert rates under configurable conditions. This tool is useful for database engineers, developers, and system administrators who want to assess MongoDB performance by simulating multiple threads performing operations in a MongoDB collection.

Features

• Configurable Load Testing: Set the number of concurrent threads and the total number of documents for testing.
• Modes of Operation:
  • Insert Mode: Inserts documents into the MongoDB collection.
  • Update Mode: Updates previously inserted documents, simulating real-world workloads with mixed read-write operations.
  • Delete Mode: Deletes existing documents from the MongoDB collection.
  • Upsert Mode: Performs upserts on documents, ensuring repeated upserts within a specified range.
  • Run-All Sequence: Runs the insert, update, delete, and upsert tests in sequence, providing a comprehensive performance assessment.
• High-Resolution Metrics: Captures and logs operation rates every second, including:
  • Total document count
  • Mean operation rate
  • m1_rate, m5_rate, m15_rate (1-minute, 5-minute, and 15-minute moving average rates)
  • Mean rate (mean_rate)
• In-Memory Logging with Final CSV Export: Stores per-second metrics in memory and exports to a CSV file after the test completes, minimizing disk I/O during the benchmark run.
• Detailed Console Output: Logs real-time performance metrics to stdout every second.

Usage

After building the tool, run it with customizable parameters:

./mongo-bench -threads <number_of_threads> -docs <number_of_documents> -uri <mongodb_uri> -type <test_type>

Parameters

• -threads: Number of concurrent threads to use for inserting, updating, deleting, or upserting documents.
• -docs: Total number of documents to process during the benchmark.
• -duration: Duration of the test in seconds (default: 0 seconds).
• -largeDocs: Use large documents (2K) (default: false).
• -dropDb: Drop the database before running the test (default: true).
• -uri: MongoDB connection URI.
• -type: Type of test to run. Accepts insert, update, delete, upsert, or runAll:
  • insert: The tool will insert new documents.
  • update: The tool will update existing documents (requires that documents have been inserted in a prior run).
  • delete: The tool will delete existing documents. (just if docs is given)
  • upsert: The tool will perform upserts, repeatedly updating a specified range. (just if docs is given)
• runAll: Runs the insert, update, delete, and upsert tests sequentially. (just if docs is given)
• runAll: Runs the insert, update tests sequentially. (just if duration is given)

Example Commands

Insert Test:

./mongo-bench -threads 10 -docs 100000 -uri mongodb://localhost:27017 -type insert

This command will insert 100,000 documents into MongoDB using 10 concurrent threads.

Update Test:

./mongo-bench -threads 10 -docs 100000 -uri mongodb://localhost:27017 -type update

This command will update the 100,000 documents previously inserted using 10 concurrent threads.

Delete Test:

./mongo-bench -threads 10 -docs 100000 -uri mongodb://localhost:27017 -type delete

This command will delete documents from MongoDB using 10 concurrent threads.

Upsert Test:

./mongo-bench -threads 10 -docs 100000 -uri mongodb://localhost:27017 -type upsert

This command will perform upserts on documents within a specified range, using 10 concurrent threads.

Run All Tests:

./mongo-bench -threads 10 -docs 100000 -uri mongodb://localhost:27017 --runAll

This command will run the insert, update, delete, and upsert tests sequentially using 10 concurrent threads.

Output

• Console: Logs per-second operation rate metrics to stdout.
• CSV: After the test completes, saves a detailed CSV file (e.g., benchmark_results_insert.csv, benchmark_results_update.csv) with columns:
  • t: Timestamp (epoch seconds)
  • count: Total document count
  • mean: Mean operation rate in docs/sec
  • m1_rate, m5_rate, m15_rate: Moving average rates over 1, 5, and 15 minutes, respectively
  • mean_rate: Cumulative mean rate

This CSV file provides an in-depth view of performance over time, which can be used for analysis or visualizations.

Example CSV Output

t,count,mean,m1_rate,m5_rate,m15_rate,mean_rate
1730906793,100000,30000.50,31000.12,30500.45,30000.25,29900.88

Building the Tool

Build the tool using the provided Makefile:

make build

License

This project is licensed under the MIT License.