Beginner’s guide to InfluxDB’s 2026 time-series platform

Moving away from Prometheus and TimescaleDB

InfluxDB IOX is the new engine powering InfluxDB Cloud, and it uses Rust and the Apache Arrow ecosystem to process time-series data. Teams that use Prometheus manage a pull-based system optimized for metrics. Prometheus graduated from the CNCF and excels in cloud-native monitoring through built-in service discovery, which helps with scalability. InfluxDB is a push-based system that suits event logs and IoT sensor data. Many teams leave TimescaleDB because they hit scaling walls with compression or struggle to manage data that lacks a time column. InfluxDB IOX addresses these issues through a columnar architecture with Apache Parquet files for on-disk storage. I recommend this platform for teams that need to manage massive amounts of high-precision event data alongside regular metrics. The project began when founders Paul Dix and Todd Persen pivoted from a SaaS company called Errplane to build an open-source database. This move followed an $8.1 million investment from Mayfield and Trinity Ventures. InfluxData later secured $16 million in Series B funding to expand its reach. Prometheus is a pull-based system that requires an application to publish metrics to an endpoint, whereas InfluxDB is a push-based system where the application sends data directly.

IOX Technical Specifications

The IOX engine allows for unbounded cardinality. Users write event data with infinite cardinality and slice data across any dimension without sacrificing performance. The columnar design uses per-column compression, dictionaries, and run length encoding to increase speed. InfluxDB IOX reaches query speeds of 1 billion rows per second per core through partitioning by time and tags. Because the new engine relies on the Apache Arrow ecosystem and DataFusion, users can now connect their time-series data to various third-party tools using standard PostgreSQL drivers instead of having to learn entirely new and proprietary communication protocols. This architecture uses DataFusion as its parser, planner, optimizer, and execution engine. Companies like Nordstrom, eBay, and Solar City use InfluxDB to manage their data, while Mozilla uses the technology to access performance metrics for the Firefox browser in real-time.

Engineering implementation and migration

I have seen engineering teams struggle with the steep learning curve of the Flux query language during previous InfluxDB iterations. Version changes often force teams to rebuild their entire visualization stack. You should plan for a shift in how you handle metadata. Previous versions of InfluxDB required many tags to manage metadata, which eventually slowed performance as cardinality grew. InfluxDB IOX removes these cardinality limits, yet you must still account for the change in data architecture.

I suggest you evaluate your current data needs before committing to a migration. Prometheus works well if you only need to monitor metrics in a dynamic Kubernetes environment. InfluxDB is a better fit if you need to store high-precision event data from sensors or IoT devices. The IOX engine is a middle ground because it provides SQL capabilities that older versions lacked. You must decide if your team can handle the move from a pull model to a push model.

The migration from Prometheus or TimescaleDB to InfluxDB IOX involves more than just changing a database. You will face new ways of handling data retention and shard groups. InfluxDB OSS v2 uses a retention enforcement service that deletes shard groups when they fall outside the bucket retention period. This service runs at intervals you can configure using the storage-retention-check-interval option. I have noticed users complain about the rollercoaster of InfluxDB version changes, specifically the shift from Go to Rust. This transition can feel like rebuilding the entire metric visualization from scratch. You should also consider that TimescaleDB is a PostgreSQL extension, which requires different management than a standalone time-series database like InfluxDB. Does the ability to run SQL queries on high-cardinality event data justify the effort of migrating your existing metric pipelines?