InfluxDB

InfluxDB est l'une des time-series databases open source les plus populaires, créée par InfluxData en 2013. Successive evolution : InfluxDB 1.x (Go, custom storage TSM), 2.x (Flux query language, OSS struggling), **3.x (2023+, totally rewritten in Rust avec DataFusion engine — massive perf improvement)**.

InfluxDB 3 (Core, Edge, Cloud Serverless) :
(1) **Apache Arrow / Parquet** — columnar format, ouvre interop avec ecosystem big data (Spark, Pandas, BI tools).
(2) **DataFusion query engine** (Apache Arrow ecosystem) — Rust SQL engine.
(3) **SQL natif** — first-class SQL alongside InfluxQL legacy (Flux deprecated).
(4) **Object storage** (S3) — separation compute/storage, infinite scalability.
(5) **High cardinality** handling — improved drastically vs v2.
(6) **Real-time + historical** unified.

Data model : (1) **Measurement** (équivalent table) ; (2) **Tags** (indexed metadata, low cardinality recommended) ; (3) **Fields** (actual values, not indexed) ; (4) **Timestamp** (primary index).

Protocols ingestion : InfluxDB line protocol (text format), HTTP API, Telegraf (Influx's collector agent), Apache Kafka.

Exemple line protocol : `cpu,host=server01,region=us-west usage_idle=89.0,usage_user=8.1 1672531200000000000`.

Query exemples (SQL InfluxDB 3) :
```sql
SELECT mean(usage_idle) AS avg_idle
FROM cpu
WHERE host = 'server01' AND time > NOW() - INTERVAL '1 hour'
GROUP BY time(5m)
ORDER BY time;
```

Use cases : (1) IoT (devices sending sensor readings) ; (2) DevOps monitoring (collected by Telegraf) ; (3) industrial sensor data ; (4) financial tick data ; (5) network/security event data.

Écosystème InfluxData : (1) **InfluxDB** TSDB ; (2) **Telegraf** — collector agent (200+ plugins, équivalent fluentd/Logstash pour metrics) ; (3) **Kapacitor** — alerting/processing (legacy, replaced by tasks Flux) ; (4) **Chronograf** (legacy UI, replaced by Influx UI).

Concurrents : TimescaleDB, Prometheus + VictoriaMetrics, QuestDB, ClickHouse, AWS Timestream, Azure Data Explorer. Compétences DEA-C01.