---
title: "Fluss White Paper"
description: "Apache Fluss® is an open-source, lakehouse-native storage layer designed to unify the fragmented infrastructure that real-time analytics, Machine Learning (ML) pipelines, and AI systems require. This paper traces the architecture that makes that possible, from the storage engine internals to the stateless compute model to the unified feature and context store. "
lastUpdated: 2026-10-14T09:40:33.000Z
source_url:
  html: "https://www.ververica.com/asset-library/fluss-white-paper"
  md: "https://www.ververica.com/asset-library/fluss-white-paper.md"
---

# Apache Fluss®: The Foundation of the Unified Streaming Lakehouse

A Lakehouse-native streaming storage for real-time analytics, context engineering, and AI.

## Apache Fluss

Apache Fluss® is an open-source, lakehouse-native storage layer designed to unify the fragmented infrastructure that real-time analytics, Machine Learning (ML) pipelines, and AI systems require.

This paper traces the architecture that makes that possible, from the storage engine internals to the stateless compute model to the unified feature and context store.

## Real-time AI systems demand infrastructure you don't have.

No organization builds a real-time AI data platform from scratch. They arrive at it through a sequence of expanding ambitions: streaming ingestion, then analytics, then event processing, then ML features, then AI systems.

Each stage adds a new specialized system. A message broker. A stream processor. An online store. An offline store. A synchronization layer. Every boundary between the systems is a seam where data silently diverges.

## Why does real-time analytics need a new storage layer?

The Apache Fluss white paper explains why real-time analytics and AI break down when the same data sits in a message broker, a stream processor, online and offline stores and a lakehouse at once. It shows how Apache Fluss, an open-source, lakehouse-native streaming storage layer, replaces those copies with one storage layer, so the synchronization problem disappears.