IoT Time-Series Database Benchmarking Framework

The Internet of Things (IoT) connects a vast network of devices that continuously generate large volumes of time-stamped data. Extracting value from such data requires storage systems optimized for high-frequency, append-only workloads. This study compares five Time-Series Databases (TSDBs) – InfluxDB, TimescaleDB, QuestDB, Apache Druid, Graphite – alongside one relational SQL system (PostgreSQL) and one NoSQL system (Cassandra). The focus is on data ingestion performance in a private cloud environment using a standardized IoT dataset.

The experiments compare relational, time-series, and NoSQL databases under identical workloads.
Benchmarks include:

• Raw ingestion and optimized ingestion codes
• Query latency across six representative queries (Q1–Q6)
• Multiple dataset sizes (5 rows, 2,000 rows, 500,000 rows)

Note: This repository only contains ingestion scripts for each database. Query benchmarking, visualization, and analysis are documented separately in the main thesis but are not part of this code repository.

Repository Structure

• The folder TimeSeriesDatabase_Bechmarks_PythonCodes contains a folders with a <Database name>, each folder contains Python scripts, configuration files, README.md, and usage instructions for benchmarking as part of the time-series database performance evaluation for IoT data.
• The folder Data has the dataset used for this benchmarking

🚀 How to Use

1. Prepare datasets – CSV files in the data/ folder.
2. Start your database service – Local or VM instance.
3. Choose the database folder and follow its README.md.
4. Run ingestion scripts to insert data.
5. Run query scripts to measure query latency.
6. Record results for comparison.

📊 Benchmarked Databases

• InfluxDB v2
• PostgreSQL
• TimescaleDB (PostgreSQL extension for time-series)
• QuestDB
• Apache Druid
• Graphite (excluded from some results due to disk usage limits)
• Apache Cassandra