G1Joshi InfluxDB
InfluxDB is a purpos-built time series database. Version 3.0 (IOx) is a complete rewrite in Rust, using Parquet/Arrow for massive performance gains.
When to Use
- IoT Metrics: Sensor data from millions of devices.
- DevOps Monitoring: CPU, RAM, Disk usage over time.
- Financial Ticks: High frequency trading data.
Quick Start (InfluxQL / SQL)
InfluxDB 3.0 supports SQL!
SELECT room, MEAN(temp)
FROM sensors
WHERE time > now() - 1h
GROUP BY room
Core Concepts
Push vs Pull
InfluxDB is "Push" based (Telegraf agents push data to it). Prometheus is "Pull" based.
High Cardinality (v3)
The old InfluxDB struggled if you had too many "Tags" (Cardinality). The new IOx engine (Parquet-based) handles unlimited cardinality.
Downsampling
Automatically aggregating high-resolution data (1s) into lower resolution (1m, 1h) to save space.
Best Practices (2025)
Do:
- Use SQL: InfluxDB 3.0 prioritizes SQL (FlightSQL) over the old Flux language.
- Use Parquet: Understand that data is stored in Parquet (Object Storage friendly).
- Tag wisely: Even though v3 handles cardinality, proper schema design (Measurement vs Tag vs Field) still matters for query speed.
Don't:
- Don't use Flux for new projects: It is being deprecated in favor of SQL and Python.