Introducing Tansu.io -- Rethinking Kafka for Lean Operations

Abstract

Tansu is an open-source, Apache Kafka®-compatible messaging broker designed to be simpler and more flexible than traditional Kafka clusters. It implements the Kafka protocol so Kafka clients and tools can work with it, but it’s built with stateless brokers and pluggable storage backends rather than relying on Kafka’s built-in replicated storage and consensus protocols.

Key Takeaways:

  • Kafka API compatible: Works with standard Kafka clients and tools (topic creation, producers, consumers).
  • Stateless brokers: Brokers don’t store persistent data themselves; they rely on external storage systems.
  • Pluggable storage engines: Supports storage in PostgreSQL, SQLite (libSQL), S3 (or S3-compatible), or in-memory for different use cases.
  • Schema validation: Messages can be validated against Avro, JSON, or Protocol Buffers schemas.
  • Open table format support: Can write topics directly into Apache Iceberg or Delta Lake formats for analytics pipelines.
  • Single binary & CLI: Distributed as one executable with command-line tools for managing topics and producing/consuming messages.
  • Simplicity: Designed to reduce operational complexity (no ZooKeeper/Raft, easy to deploy and scale) while enabling lightweight Kafka-compatible streaming.

In short, Tansu is like a lightweight, Kafka-compatible streaming platform built for simplicity, flexible storage, and modern data pipelines.


Speaker

Peter Morgan

Founder @tansu.io

Peter is the founder of tansu.io, an open-source, Apache Kafka®-compatible broker designed around simplicity and statelessness. Working with a growing community of contributors, he is reimagining messaging infrastructure through single-binary deployment, built-in schema validation, and pluggable storage backends.

A programmer since the ZX81 era, Peter has spent his career building network protocols and distributed, event-driven systems across multiple industries—nearly always with Apache Kafka at the core. At tansu.io, he is exploring what becomes possible when compute is fully decoupled from storage and broker state is eliminated altogether.

Some links: 

Read more
Find Peter Morgan at:

Date

Tuesday Mar 17 / 10:35AM GMT ( 50 minutes )

Location

Fleming (3rd Fl.)

Share

From the same track

Session

From S3 to GPU in One Copy: Rethinking Data Loading for ML Training

Tuesday Mar 17 / 01:35PM GMT

ML training pipelines treat data as static. Teams spend weeks preprocessing datasets into WebDataset or TFRecords, and when they want to experiment with curriculum learning or data mixing, they reprocess everything from scratch.

Speaker image - Onur Satici

Onur Satici

Staff Engineer @SpiralDB

Session

The Rise of the Streamhouse: Idea, Trade-Offs, and Evolution

Tuesday Mar 17 / 11:45AM GMT

Over the last decade, streaming architectures have largely been built around topic-centric primitives—logs, streams, and event pipelines—then stitched together with databases, caches, OLAP engines, and (increasingly) new serving systems.

Speaker image - Giannis Polyzos

Giannis Polyzos

Principal Streaming Architect @Ververica

Speaker image - Anton Borisov

Anton Borisov

Principal Data Architect @Fresha

Session

Ontology‐Driven Observability: Building the E2E Knowledge Graph at Netflix Scale

Tuesday Mar 17 / 02:45PM GMT

As Netflix scales hundreds of client platforms, microservices, and infrastructure components, correlating user experience with system performance has become a hard data problem, not just an observability one.

Speaker image - Prasanna Vijayanathan

Prasanna Vijayanathan

Engineer @Netflix

Speaker image - Renzo  Sanchez-Silva

Renzo Sanchez-Silva

Engineer @Netflix

Session

Building a Control Plane for Production AI

Tuesday Mar 17 / 03:55PM GMT

Details coming soon.

Session

Unconference: Modern Data Engineering

Tuesday Mar 17 / 05:05PM GMT