whichdb

Apache Fluss

active

by Apache Software Foundation · since 2024 · also known as Fluss

Columnar streaming storage for Flink with primary-key tables and tiering into the lakehouse.

columnar / olapkey-valuedatabase engineopen source (permissive)emerging
Website Docs Source★ 2.2K Compare Find similar

Overview

Fluss is a streaming storage layer designed for real-time analytics with Apache Flink, positioned as a table-oriented alternative to Kafka for analytical pipelines. Log tables store append-only streams in an Arrow-based columnar format with projection pushdown, and primary-key tables support upserts, changelog generation and low-latency key lookups for lookup joins. A tiering service continuously compacts data into Paimon, Iceberg, Hudi or Lance so streaming and lakehouse reads can be unioned. Open-sourced by Alibaba in late 2024, it entered the Apache Incubator in 2025 and is now a top-level project; 1.0.0 shipped in September 2026.

Stream processingReal-time analyticsLakehouse

Strengths & trade-offs

Strengths

  • +Columnar stream storage lets Flink read only needed columns
  • +Primary-key tables replace external KV stores for lookup joins
  • +Built-in tiering into Paimon, Iceberg, Hudi and Lance
  • +Apache governance with Alibaba-scale origins

Trade-offs

  • −Young project; 1.0 only in 2026
  • −Best used with Flink; other engines are secondary
  • −Not a general Kafka replacement for non-analytical messaging

Great for

  • +Real-time analytics pipelines on Flink
  • +Streaming lakehouse ingestion with fresh reads
  • +Lookup joins against changing dimension tables

Think twice for

  • −Applications without Flink or Spark
  • −General-purpose message queuing

Capabilities

Native or official first-party support · 7 of 25 tracked

✓Secondary indexes
✓Change data capture
✓TTL / expiry
✓Columnar storage
✓Durable persistence
✓Compression
✓Streaming ingest
–Transactions
–Joins
–Full-text search
–Vector search
–Geospatial
–JSON
–Triggers
–Stored procedures
–Materialized views
–Pub/Sub
–In-memory
–Encryption at rest
–Role-based access
–Time travel
–Branching
–Online schema changes
–Offline sync
–Runs in WebAssembly

Scale & consistency

Comfortable data size
Large · ≤ 1 PB
Horizontal scaling
Native
Multi-region
No
Transactions
Limited ACID
Consistency
Strong
Schema
Strict
Architecture
Distributed, Client server
Query languages
Flink SQL, Spark SQL

Tablet servers hold bucketed tables with replication; cold data tiers to lake storage.

Deployment

Self-hosted Vendor-managed cloud Serverless Embedded Edge Kubernetes operator

Ecosystem

apis & protocolsJava client, REST gateway
client languagesJava
used byAlibaba

Alternatives to Apache Fluss

Compare all→
RisingWaveRelational17.4

An Apache-licensed, Postgres-compatible streaming database that stores state on S3 and writes Iceberg.

apache-fluss vs risingwave →
Apache PaimonColumnar / OLAP14.2

Streaming lakehouse table format with LSM-based upserts and changelogs, born in Apache Flink.

apache-fluss vs apache-paimon →
MaterializeRelational15.7

A Postgres-compatible operational data warehouse that keeps SQL views incrementally up to date.

apache-fluss vs materialize →
Timeplus ProtonColumnar / OLAP16.6

A single C++ binary for streaming SQL, built on ClickHouse's engine with streams and windows.

apache-fluss vs timeplus-proton →
ArroyoRelational5.4

A Rust stream processor with SQL pipelines, now owned by Cloudflare and powering Cloudflare Pipelines.

apache-fluss vs arroyo →
FelderaRelational15.8

An incremental SQL engine grounded in DBSP theory that updates any SQL query from changes only.

apache-fluss vs feldera →

Apache Fluss FAQ

What is Apache Fluss?

Apache Fluss is a columnar / OLAP database by Apache Software Foundation, first released in 2024. Fluss is a streaming storage layer designed for real-time analytics with Apache Flink, positioned as a table-oriented alternative to Kafka for analytical pipelines. Log tables store append-only streams in an Arrow-based columnar format with projection pushdown, and primary-key tables support upserts, changelog generation and low-latency key lookups for lookup joins. A tiering service continuously compacts data into Paimon, Iceberg, Hudi or Lance so streaming and lakehouse reads can be unioned. Open-sourced by Alibaba in late 2024, it entered the Apache Incubator in 2025 and is now a top-level project; 1.0.0 shipped in September 2026.

Is Apache Fluss open source?

Yes. Apache Fluss is open source. License: Apache License 2.0.

Does Apache Fluss support ACID transactions?

Partly. Apache Fluss offers limited ACID guarantees (for example, single-record or single-partition transactions), with strong consistency.

Does Apache Fluss scale horizontally?

Apache Fluss scales horizontally out of the box, with built-in sharding or partitioning. It is comfortable at large scale (≤ 1 PB).

How can I run Apache Fluss?

You can self-host it.

What is Apache Fluss best for?

Apache Fluss is a strong fit for real-time analytics pipelines on Flink, streaming lakehouse ingestion with fresh reads and lookup joins against changing dimension tables. Think twice for applications without Flink or Spark.

What are the best alternatives to Apache Fluss?

Common alternatives to Apache Fluss include RisingWave, Apache Paimon, Materialize, Timeplus Proton and Arroyo. The right choice depends on your workload, scale and hosting; WhichDB's smart match ranks them against your requirements.

How popular is Apache Fluss?

Apache Fluss ranks #323 of 844 databases on the WhichDB Score (8.4/100), and #31 among columnar / OLAP databases. The score combines public adoption signals such as package downloads, GitHub activity, job posts and developer discussion.