Managed Time Series Databases (TSDB)

Why it exists

General-purpose databases like PostgreSQL or MongoDB are not built for the unique characteristics of time-series data: massive write volumes, data that is rarely updated, and queries that are almost always filtered and aggregated by time. Using them for this purpose leads to slow queries, inefficient storage, and operational headaches as data volume grows.

The mental model

A managed time series database is like a specialized warehouse-as-a-service for time-stamped packages. You don't build the warehouse or manage the staff; you just send your packages (data points), and the service handles sorting them by time, storing them efficiently, and letting you retrieve them instantly, no matter how many arrive per second.

How it works

Managed TSDBs are optimized for time as the primary dimension. They ingest huge streams of data points, each with a timestamp and associated metadata (dimensions). Under the hood, they use techniques like time-based partitioning, columnar storage for better compression, and automatic data tiering. Recent, frequently queried data is kept in fast, expensive storage (like RAM), while older data is automatically moved to cheaper, long-term storage, all while remaining queryable. The 'managed' part means the cloud provider handles all this complexity, plus scaling, replication, and backups.

When to use it

Use a managed TSDB when you are collecting high-volume measurements over time. Three common use cases are: first, observability, for collecting application and infrastructure metrics, logs, and traces; second, IoT, for ingesting data from fleets of sensors and devices; third, financial services, for tracking market data and trade history.

When not to use it

Avoid TSDBs for highly relational or transactional data, such as user profiles, product catalogs, or e-commerce orders. This data has complex relationships and frequent updates, which is the exact workload that relational databases like PostgreSQL or MySQL are designed to handle well.

One canonical example

Amazon Timestream is a serverless managed time series database. It automatically scales to handle high-throughput writes and complex queries. It features an adaptive query engine and separates storage into a fast in-memory tier for recent data and a cost-effective magnetic tier for historical data, providing a single interface to query across both.