Best Time-Series Databases for IoT and Telemetry

TimescaleDB

TimescaleDB, from Tiger Data, is a PostgreSQL extension for real-time analytics on time-series and event data. Hypertables partition tables by time into chunks automatically; the hypercore engine keeps recent data in a rowstore and converts older chunks to a compressed columnstore; continuous aggregates refresh rollups incrementally; and retention policies drop whole chunks. Because it is PostgreSQL, readings can be joined to asset registers and CMMS work orders in one query. The latest release on GitHub when checked was 2.31.0 (20 October 2026).

Licence: the repository's LICENSE file states that outside the tsl directory "source code in a given file is licensed under the Apache License Version 2.0, unless otherwise noted", and that within it source code "is licensed under the Timescale License, unless otherwise noted". Tiger Data's edition comparison places the columnstore, continuous aggregates, retention policies and the job scheduler in Community Edition (TSL). Tiger Data's licence page summarises the TSL as: "Roughly speaking, as long as you are not offering TimescaleDB as a hosted Database-as-a-Service, you can use all Community features for free." The TSL agreement is more detailed: it grants Internal Use, with conditions on who may reach the database's data definition and manipulation interfaces, and a separate Value Added Products or Services licence, and its section 2.2 prohibits "using any TSL Licensed Software to provide time-sharing services or database-as-a-service services". See the TimescaleDB guide and TimescaleDB vs PostgreSQL.

Official site

InfluxDB 3

InfluxDB 3 is written in Rust on Apache Arrow, DataFusion and Parquet. Data is organised in databases and tables whose columns are time, tags (identifiers such as site or sensor ID) and fields (values); it is written as line protocol and queried with SQL or InfluxQL. Flux, the language of InfluxDB 2.x, is not part of InfluxDB 3. An embedded Python processing engine runs plugins for transforms, downsampling and alerting, and the plugin library includes MQTT and Kafka subscribers. The vendor reports query response times under 10 ms for last-value queries. The latest release on GitHub when checked was v3.12.1 (dated 14 October 2026 in the release list).

Licence: the InfluxDB repository states: "The open source software we build is licensed under the permissive MIT or Apache 2 licenses at the user's choosing." (texts in LICENSE-MIT and LICENSE-APACHE). InfluxDB 3 Core is a single-node engine; InfluxData describes it as purpose-built for real-time monitoring and recent data. InfluxDB 3 Enterprise runs under a licence from InfluxData (its licence page lists Trial, At-Home and Commercial licences) and adds high availability, multi-node clustering, read replicas and long-range compaction. In Core, a database's retention period is set at creation and cannot be changed later. See the InfluxDB guide and TimescaleDB vs InfluxDB.

Official site

QuestDB

QuestDB describes itself as an open source, low-latency time-series database on open formats. Data lives in time-partitioned columns; older partitions can be converted to Apache Parquet. Queries use SQL with time-series extensions: SAMPLE BY groups rows into time buckets and requires a designated timestamp column. Clients can query over the PostgreSQL wire protocol with standard PostgreSQL libraries. For ingestion, QuestDB now recommends its own QuestDB Wire Protocol (QWP) for new clients and keeps InfluxDB line protocol as a compatibility path for Telegraf and existing pipelines. The latest release on GitHub when checked was 10.1.0 (22 October 2026).

-- Illustrative, not run here: from the QuestDB documentation; check it against the
-- documentation for your version before use.
SELECT ts, count() FROM trades SAMPLE BY 1h;

Licence: the repository's LICENSE.txt is the Apache License 2.0. QuestDB Enterprise is commercial and adds replication with automatic failover, automatic tiering of cold partitions to object storage, and SSO, role-based access control and audit logs. See the QuestDB guide.

Official site

VictoriaMetrics

VictoriaMetrics is a time-series database and monitoring system. It implements MetricsQL, a PromQL-like query language, and accepts data through Prometheus scraping and remote write, InfluxDB line protocol, Graphite, OpenTSDB, OpenTelemetry, CSV and several agent formats, which lets it serve as a long-term store behind Prometheus. Retention is a global setting that defaults to one month (31 days). The latest release on GitHub when checked was v1.155.0 (26 October 2026).

Licence: the repository states that the single-node and cluster versions are under Apache License 2.0 (text in its LICENSE file). VictoriaMetrics Enterprise is commercial and adds downsampling, multiple retentions via retention filters, and long-term support releases. In our view it fits BMS data when readings are modelled as metrics with labels (site, building, point), and fits less well when you need SQL joins to business data.

Official site

TDengine

TDengine TSDB is a time-series database that its documentation describes as designed for IoT platforms, industrial internet, power and IT operations. Data model: one table per data collection point. A supertable is the template for one device type: it defines the timestamp, the metric columns and the tags (static attributes such as location, model or group) and stores no data itself. Each device gets a subtable created from that template with its own tag values, and a query on the supertable covers all of its subtables, filtered by tag. Virtual tables combine columns from several source tables for devices whose sensors report at different rates.

Query language: standard SQL with time-series extensions: windows (INTERVAL, SESSION, STATE_WINDOW, EVENT_WINDOW, COUNT_WINDOW) with FILL, PARTITION BY tbname to compute each device separately, and functions such as time-weighted average; JDBC, ODBC, REST and client libraries connect BI tools and applications. Ingestion: SQL INSERT, which can create a device's subtable on first write, or schemaless writes in InfluxDB line protocol, OpenTSDB telnet or OpenTSDB JSON, which create supertables and subtables automatically. The zero-code data connectors for MQTT, OPC UA, OPC DA, Kafka, PI System and other sources run in taosX and are configured in taosExplorer; the documentation marks them, and taosX itself, as available in TDengine TSDB-Enterprise only, so with TSDB-OSS a bridge such as Telegraf or your own subscriber has to move data from an MQTT broker.

Retention, caching and processing: retention is set per database with KEEP (default 3,650 days, at least three times the file span DURATION); the architecture documentation says expired data files are deleted automatically. Multi-tier storage with several KEEP values and RSMA downsampled storage are Enterprise features; the community version applies the longest KEEP. CACHEMODEL (last_row, last_value or both) keeps the latest row or values of each subtable in memory so LAST_ROW and LAST serve current-status dashboards without reading history, at some cost on the write path. Built-in stream processing defines real-time transformations in SQL and writes results to target tables, and data subscription publishes topics through a Kafka-like API.

Deployment: clusters are deployed manually, with Docker or on Kubernetes; a highly available cluster needs three or more dnodes, three mnodes and databases with three replicas. TSDB-Enterprise can run at the edge, on premises or in public or private clouds, and TDengine Cloud delivers the Enterprise feature set as a managed service on AWS, Azure and Google Cloud. The latest release on GitHub when checked was 3.4.3.3 (release date in its notes: 18 October 2026); the documentation's release notes already list 3.4.3.6 (24 October 2026). Architecture, a plant-room schema and when not to use it are in the TDengine guide.

Licence: TDengine TSDB-OSS is under the GNU Affero General Public License v3.0 (AGPL-3.0); the vendor states that it includes clustering and is the code base of its paid offerings. AGPL-3.0 is a copyleft licence. Section 13 of the AGPL-3.0 text, "Remote Network Interaction", says: "Notwithstanding any other provision of this License, if you modify the Program, your modified version must prominently offer all users interacting with it remotely through a computer network (if your version supports such interaction) an opportunity to receive the Corresponding Source of your version". How that applies to a given deployment or product depends on the full licence text and the facts; this is not legal advice, so read the licence and take legal advice for your situation. TSDB-Enterprise adds, among other things, the zero-code connectors, fine-grained permissions, IP whitelisting, audit logs and encryption in transit and at rest. We did not confirm Enterprise or Cloud prices on a vendor pricing page, so none are quoted.

Official site

Apache IoTDB

Apache IoTDB is an Apache Software Foundation project for collecting, storing and analysing time-series data, aimed at industrial IoT. It stores data in TsFile, a columnar file format designed for time series. Version 2.0 offers two models: a tree model, where each data point is a path such as root.db.device01.metric that mirrors the physical plant, and a table model; the SQL dialect is chosen per session. It includes a built-in MQTT service, so devices can write directly without a separate broker, alongside native APIs for Java, Python, C++, Go and others, JDBC and REST. The latest release on GitHub when checked was 2.0.11 (11 September 2026).

Licence: the repository's LICENSE file is the Apache License 2.0. The built-in MQTT service is convenient for small deployments; at larger scale, in our view, a dedicated broker or Kafka between devices and the database still separates streaming from storage more cleanly.

Official site

ClickHouse

ClickHouse is an open source, column-oriented database for analytical reports, under Apache 2.0. Its MergeTree tables sort by a key you choose, partition by time and use TTL rules to delete or move data by age, and its documentation has a dedicated time-series guide. It is not a dedicated time-series database: there is no device model or last-value cache, primary keys do not enforce uniqueness, and updates work differently from an OLTP database. In our view it fits when the main job is aggregating very large numbers of readings across many sites, for example as the analytics layer fed from Kafka alongside an operational store.

See ClickHouse vs PostgreSQL for the model differences.

Official site

Prometheus

Prometheus, a Cloud Native Computing Foundation project under Apache 2.0, is a monitoring system: it scrapes metrics over HTTP (a pull model), stores them in a local time-series database, and evaluates PromQL rules for alerting. Its documentation is clear about the limits: local storage is not clustered or replicated, retention defaults to 15 days unless configured, and it is not a good choice where you need 100% accuracy such as per-request billing. The latest release on GitHub when checked was v3.15.1 (21 October 2026).

For BMS and IoT, Prometheus suits monitoring the gateways, brokers and servers in the pipeline. Our editorial view: for the sensor readings themselves, which you may need for years and for energy or compliance reports, use a database built for retention and SQL, or send Prometheus data by remote write to a long-term store such as VictoriaMetrics. See Prometheus vs Datadog.

Official site

Plain PostgreSQL

Before adopting a specialist database, check whether PostgreSQL alone is enough. Declarative range partitioning by time, a BRIN index on the timestamp, date_bin for bucketing, materialized views for rollups and the pg_partman extension for partition maintenance and retention can be enough at modest volumes (as an illustrative scale, hundreds to a few thousand points at minute intervals; where the limit lies depends on your hardware, query patterns and retention, so test with your own data), and keep readings next to assets and work orders. What it lacks is columnar compression and incrementally refreshed rollups; that is where TimescaleDB, an extension to the same database, comes in.

Licence: the PostgreSQL License, which the project describes as "a liberal Open Source license, similar to the BSD or MIT licenses"; pg_partman's LICENSE.txt states that it is released under the same licence. Examples that were run on PostgreSQL 18.6 are in TimescaleDB vs PostgreSQL; to set up a server, see PostgreSQL in Docker.

Official site