NoSQL Analytics Optimization: MongoDB vs Redis

Overview

This project compares MongoDB and Redis for analytical workloads using a large-scale e-commerce dataset (~2M+ records). It evaluates how each system performs when executing aggregation queries on transactional data.

Key Features

• Data ingestion pipelines for MongoDB and Redis

• Data cleaning and normalization process

• Analytical queries on:

  • Top-selling category
  • Brand with highest revenue
  • Month with most sales
  • Execution time benchmarking

Tech Stack

• Python
• MongoDB
• Redis
• Docker

Dataset

E-commerce purchase history dataset with ~2.6M records, reduced to ~2.0M after cleaning.

Methodology

1. Load cleaned data into MongoDB (document model)
2. Load transformed data into Redis (key-value model)
3. Execute equivalent queries in both systems
4. Measure execution time for each query

Results

Query	MongoDB	Redis
Top Category	~5.9 sec	~11 min
Top Brand (Revenue)	~5.8 sec	~12 min
Top Month	~3.7 sec	~12 min

Key Insights

• MongoDB outperforms Redis by up to 100x+ in analytical queries
• Redis is inefficient for aggregation-heavy workloads due to lack of query engine
• MongoDB’s aggregation pipeline and indexing enable scalable analytics
• Data cleaning was critical to obtain valid results

How to Run

docker-compose up
pip install -r requirements.txt

Load data:

python src/ingestion/load_mongo.py
python src/ingestion/load_redis.py

Run queries:

python src/queries/mongo/query_mongo_brand.py
python src/queries/redis/query_redis_brand.py

Conclusion

MongoDB is significantly more suitable for large-scale analytical queries, while Redis is better suited for caching and fast key-value access.

Limitations

• Redis queries are executed client-side (Python), not natively
• No parallelization implemented

Author

Edgar Antonio Zeledón Pérez