TSDBs

Time series databases TSDBs are specialized databases optimized for storing, indexing, and querying time-stamped data. They are designed to support high ingest rates, long-running series, and efficient retrieval of data over time ranges, often with built-in mechanisms for downsampling, retention, and compression.

Data model: A TSDB stores data points identified by a timestamp, a metric or measurement name, and one or more tags or labels that describe the series. Each point carries one or more numeric or boolean fields. The combination of measurement, tags, and timestamp defines a unique series.

Storage and indexing: TSDBs typically use an append-only storage model with time-based partitioning into segments or shards. Data is compressed with time-series specific encodings and may be stored columnar in parts. Indexes usually cover time and tag values to accelerate range scans, filters, and aggregations.

Ingestion and queries: Ingestion aims for high throughput with lightweight write paths. Retention policies and downsampling support long-term storage while reducing resolution. Queries commonly specify a time range, apply aggregations (sum, avg, max, min, percentile), and filter by tag values. Some systems support continuous queries or rollups.

Use cases: Common applications include monitoring and observability, IoT telemetry, financial tick data, and any scenario requiring metric-like time series data and fast aggregations over time.

Architecture and scalability: Many TSDBs offer clustering, sharding, data compaction strategies, and tiered storage. Trade-offs involve consistency guarantees, durability, and complexity of querying across distributed partitions.

Examples: Popular TSDBs include InfluxDB, Prometheus, TimescaleDB, OpenTSDB, KairosDB, and VictoriaMetrics. Each has different ingestion models, query languages, and deployment options.