Cloud Parts

InfluxDB

The best-known time-series database, fed by line protocol and Telegraf; now in its third generation, a Rust rewrite that keeps data as Parquet files on object storage.

What it is
Time-series databases
Type
Open source
Made by
InfluxData Inc.
License
MIT or Apache 2.0 (InfluxDB 3 Core); MIT (1.x and 2.x); InfluxDB 3 Enterprise and InfluxDB Cloud are proprietary
Facts checked
6 October 2026
Website Source code

What it is

InfluxDB stores measurements (tables) of timestamped points with tags and fields, written as line protocol over HTTP, usually by Telegraf or a client library. It exists in three incompatible generations: 1.x and 2.x are a Go server on InfluxData’s own Time-Structured Merge Tree (TSM) engine with a series index, and 3.x, the current one, is a Rust rewrite on Apache Arrow, DataFusion and Parquet. In InfluxDB 3 Core, one influxdb3 process on port 8181 validates each write into memory, flushes a write-ahead log to object storage (S3, Google Cloud Storage, Azure Blob or local disk) every second, keeps recent data in a queryable buffer, and every ten minutes persists the oldest data as Parquet files. Queries use SQL or InfluxQL over HTTP or Arrow Flight SQL, helped by last-value and distinct-value caches, and an embedded Python processing engine runs plugins on new writes or on a schedule.

Core is free under MIT or Apache 2.0 but has no compactor: its Parquet files stay in small 10-minute blocks, and by default a query may read at most 432 of them, about 72 hours of data. InfluxDB 3 Enterprise, under a commercial licence priced by CPU cores, adds compaction for long-range queries, multi-node clusters with high availability and read replicas, and SSO with role-based access; a free At-Home licence allows two CPUs for non-commercial use. Managed options are InfluxDB Cloud Serverless (usage billing), Cloud Dedicated (single-tenant, custom price) and Amazon Timestream for InfluxDB. InfluxDB 1.x and 2.x are still MIT-licensed and still get releases, 1.13.1 and 2.9.1 in 2026.

Why teams pick it

  • The biggest ecosystem in the category: Telegraf’s 400+ plugins, client libraries, and a data source preinstalled in Grafana that speaks SQL, InfluxQL and Flux
  • Line protocol is simple and widely spoken, and InfluxDB 3 still accepts the 1.x and 2.x write APIs and InfluxQL, so existing agents and many dashboards keep working
  • InfluxDB 3 keeps data as Parquet on object storage, without the series-cardinality limits of the old index, in files other tools can read
  • Core is permissively licensed and starts as one binary with no dependencies, with a Python engine for downsampling, alerting and transforms
  • Many ways to run it: self-hosted Core or Enterprise, InfluxData’s Cloud Serverless or Dedicated, or Amazon Timestream for InfluxDB on your AWS bill

Watch out for

  • Core is for recent data: with no compactor, a query reads at most 432 Parquet files by default (about 72 hours); raising the limit costs memory and object-store requests, and long-range history is an Enterprise feature
  • Three generations, three engines: Flux (2.x) is in maintenance mode and absent from InfluxDB 3, TSM data must be migrated, and since 15 September 2026 the Docker latest tag points to InfluxDB 3 Core, so unpinned 1.x or 2.x containers can jump versions
  • Clustering and high availability have been closed source since 2016; InfluxDB 3 Enterprise is priced per CPU core by quote, and the free At-Home licence is limited to two CPUs and non-commercial use
  • The 2023 plan promised a free InfluxDB 3 “Community” edition with a compactor; what shipped in 2025 was Core without one, plus the non-commercial At-Home tier
  • Cloud Serverless bills on four meters, including $0.09 per GB of data out, and has service quotas and no high availability

Alternatives

What all time-series databases have in common →
Amazon TimestreamAWSAWS’s time-series databases: a serverless engine now closed to new customers, and managed InfluxDB 2 and 3.Apache IoTDBOpen sourceAn Apache time-series database for industrial IoT that stores everything in its own open file format, TsFile, and lets you model devices as a tree of paths or as SQL tables.+1 TimechoDBArcOpen sourceA young single-binary time-series database in Go that writes incoming data as Parquet files and queries them with DuckDB SQL.ArcticDBCommercialMan Group’s serverless DataFrame database: a Python library with a C++ engine that stores versioned pandas DataFrames directly in S3, Azure Blob Storage or local LMDB.GreptimeDBOpen sourceA Rust database that stores metrics, logs and traces as tables on object storage, queried with SQL or PromQL.kdb+CommercialA columnar time-series database that is also an interpreter for q, an array language, used for market data across trading and risk.QuestDBOpen sourceA columnar time-series database in Java, C++ and Rust that ingests millions of rows a second and answers time-series SQL, with old partitions kept as Parquet.TDengineOpen sourceA clustered time-series database for IoT and industrial data that gives every device its own table, grouped under “supertables” with tags.TimescaleDBOpen sourceA PostgreSQL extension that turns ordinary tables into time-partitioned, compressed time-series tables.Workers Analytics EngineCommercialCloudflare’s time-series analytics store for Workers: a non-blocking `writeDataPoint()` call to record events with unlimited cardinality, adaptive sampling to keep it cheap, and a SQL API to query them.

How it fits with other technologies

InfluxDB 3 Core on S3: write line protocol, query with SQL or InfluxQL · bash

# one Core node that keeps its WAL and Parquet files in an S3 bucket
influxdb3 serve \
  --node-id host01 \
  --object-store s3 \
  --bucket my-influxdb-bucket \
  --aws-default-region us-east-1

# in another shell: create an admin token, then use it for the commands below
influxdb3 create token --admin
export INFLUXDB3_AUTH_TOKEN='<the admin token printed above>'

influxdb3 create database --retention-period 30d sensors

# line protocol: measurement,tags fields timestamp
influxdb3 write --database sensors --precision s \
  'home,room=Kitchen temp=22.7,hum=36.0,co=9i 1641056400
home,room=Kitchen temp=23.3,hum=36.9,co=18i 1641060000'

# SQL (DataFusion) or InfluxQL over the same data
influxdb3 query --database sensors \
  "SELECT date_bin(INTERVAL '1 hour', time) AS hour, room, avg(temp) AS avg_temp
   FROM home GROUP BY 1, room ORDER BY 1"
influxdb3 query --database sensors --language influxql \
  "SELECT mean(temp) FROM home GROUP BY time(1h), room"

Three generations of InfluxDB#

The name stayed the same, but the engine, query language and licence model changed. All three are still released; only 3.x gets new features.

1.x2.x3.x
General availability1.0 in September 20162.0 in November 2020On InfluxData’s cloud by 2023; Core and Enterprise in April 2025
Written inGoGoRust
StorageTSM files plus a series index (in memory or TSI) on local diskTSM files and TSI on local diskWAL and Parquet files on object storage or local disk
Query languagesInfluxQL; Flux optionalFlux and InfluxQLSQL and InfluxQL; no Flux
Open-source licenceMITMITCore: MIT or Apache 2.0; Enterprise is proprietary
ClusteringClosed-source InfluxDB Enterprise since 2016None in open sourceInfluxDB 3 Enterprise only
Status in 2026Maintained; 1.13.1 in SeptemberMaintained; 2.9.1 in May; Flux in maintenance modeCurrent; 3.12 in October
  • InfluxDB 3 accepts writes in the 1.x and 2.x formats and answers InfluxQL, but Flux queries and tasks must be rewritten, and TSM data has to be exported and written again.

InfluxDB 3 editions and pricing#

List prices checked on 2026-10-06. Amazon Timestream for InfluxDB, run by AWS on its own consumption pricing, is a fifth option.

FeatureCoreEnterpriseCloud ServerlessCloud Dedicated
Who runs itYouYou: on-prem, private cloud or edgeInfluxData, multi-tenantInfluxData, single-tenant
PriceFreeCustom, by CPU cores, bought in blocks of 8 to 128; 30-day trial; free At-Home licence (2 CPUs, non-commercial)$0.0025 per MB written, $0.012 per 100 queries, $0.002 per GB-hour stored, $0.09 per GB out; $250 starting creditCustom, by CPU, RAM and storage
Nodes and availabilityOne node, no HAMulti-node, HA, read replicasNo HA; service quotasHA, read replicas, no quotas
Sign-in and accessTokensTokens, passwords, SSO, RBACNot stated on the pricing pageSAML/SSO as an add-on
Default limits5 databases, 2,000 tables (configurable)100 databases, 10,000 tables (configurable)Service quotasNo service quotas
  • Core has no compactor and reads at most 432 Parquet files per query by default, about 72 hours of data. Enterprise, including the At-Home licence, compacts files for historical queries.

See it in a real stack

No architecture uses InfluxDB yet. Drop it into the playground and wire it up yourself.Open playground →

Benchmarks

Benchmark · Tiger Data (formerly Timescale)Time Series Benchmark Suite (TSBS)A Go toolkit that generates synthetic server-monitoring or IoT time series, bulk-loads it into a database and times a fixed set of time-series queries.Includes InfluxDB · 6 caveats to read first
How to read a benchmark →
Where it came from2013–2026

In 2011 Paul Dix and John Shahid worked at a fintech startup tracking hundreds of thousands of financial instruments, and found OpenTSDB and Graphite inadequate; Dix felt he had now solved the same time-series problem twice. Their next company, Errplane, went through Y Combinator’s Winter 2013 batch selling a monitoring service. At the Monitorama conference in 2013 they found validation not for a new monitoring product but for an idea they had long discussed, an open-source time-series database, and released InfluxDB that autumn.

Errplane renamed itself InfluxData in late 2015, the year the team replaced its storage layer with its own Time-Structured Merge Tree engine. In 2016 it moved clustering into the closed-source Enterprise edition, noting that fewer than half a percent of users ran clusters. 1.0 shipped in September 2016, and 2.0 in 2020 folded the rest of the TICK stack into one binary with a new language, Flux.

But the index behind every series grew with each new tag value, and memory-mapped files suited neither containers nor object storage. In November 2020 Dix announced IOx, a Rust rewrite on Arrow, DataFusion and Parquet. It became InfluxDB 3: first in InfluxData’s cloud, then, in April 2025, as the open-source Core and the commercial Enterprise.

Why it’s built this way

No external dependencies and a SQL-like language
The 2013 site promised a distributed database with no external dependencies, an HTTP API that JavaScript could call directly, and a SQL-like query language for time series, so developers could start without assembling a stack.
Its own storage engine, the TSM tree
General-purpose engines weren’t built for time series. TSM borrowed the LSM pattern (a write-ahead log, read-only files, compactions) and compressed each column by type, using up to 98% less disk than the engine before it.
Clustering only in the paid edition
Support contracts alone couldn’t fund development, and fewer than half a percent of users ran active clusters, so InfluxData kept the standalone server open and sold clustering and high availability.
Rebuild on Arrow, DataFusion and Parquet
Every new tag value created a series that had to be indexed, so high-cardinality data could make the index bigger than the data. Columnar Parquet files on object storage removed that index and separated compute from storage; Rust gave finer control over runtime behaviour.

Timeline

  1. 2013

    Errplane releases InfluxDB v0.0.1 in October

  2. 2015

    TSM storage engine; Errplane becomes InfluxData

  3. 2016

    Clustering goes closed source; 1.0 in September

  4. 2020

    2.0 with Flux in November; IOx rewrite announced

  5. 2023

    Flux put in maintenance mode; IOx code moves into the InfluxDB repo

  6. 2025

    InfluxDB 3 Core and Enterprise generally available in April

  7. 2026

    Docker latest tag switches to InfluxDB 3 Core

Sources24 · facts checked 6 October 2026

Improve this page on GitHub