Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Fluss 0.7 Webinar

Fluss 0.7 Webinar

Co-authored with Mehul Batra and Giannis Polyzos

Video: https://www.youtube.com/watch?si=4LFSW45GehWSIj_9&v=G-1yiYN2qoo&feature=youtu.be

Avatar for Jark Wu

Jark Wu

June 24, 2025
Tweet

More Decks by Jark Wu

Other Decks in Technology

Transcript

  1. © 2025 Ververica Thanks to all the contributors Benchao Li,

    CaoZhen, Feng Wang, Giannis Polyzos, HZY, Hongshun Wang, Jark Wu, Junbo wang, Kerwin, Leonard Xu, MehulBatra, Michael Koepf, Min Zhao, Nicholas Jiang, Radek Grebski, Rohan Dubey, Xiaojian Sun, Yang Guo, dao-jun, gkatzioura, luoyuxia, majialong, xiaozhou, yunhong, yuxia Luo, yx9o, zhangmang, 道君
  2. © 2025 Ververica Fluss v0.7: Stability • Battle-tested in Alibaba

    for massive-scale workloads ◦ Processing 10 GB/second • Production-Ready for most use cases ◦ Online/Offline Node Optimization ◦ Server Metadata Cache ◦ Rake-aware Support ◦ Accelerated Table Creation ◦ Read/Write Path Optimization
  3. © 2025 Ververica Fluss v0.7: New Streaming Lakehouse • Enhanced

    Service Robustness • Flexible Resource Management • Dynamic Task Orchestration • Actionable Offset Visibility • Single Point of Failure • Limited Scalability • Inflexible Scheduling • Opaque State Legacy Architecture New Architecture
  4. © 2025 Ververica 3-month Time-based Releases v0.5.0 OpenSource 2024.11 v0.6.0

    2025.03 v0.7.0 2025.06 v0.8.0 2025.09 Joining Apache Incubator
  5. © 2025 Ververica Lance Lakehouse https://github.com/alibaba/fluss/issues/1155 • Native ML-Optimized Storage:

    Ingest data in Lance, leveraging its columnar layout for AI/ML workloads. • High-Performance Random Access: Fast point lookups and efficient selective reads for real-time ML inference and training. Applications • Real-time AI Data Lakes • Real-time Feature Stores
  6. © 2025 Ververica Iceberg Lakehouse Effortlessly bridges Fluss streaming tables

    to Apache Iceberg tables - Analytics Ready • ETL-Free • Improved Data Freshness & Solves the Iceberg Updates Problem • Bring your own storage • Unified Data Locality via Partitioning • Union Reads • Automated Maintenance
  7. © 2025 Ververica Python Client Easiest & Lightest Route from

    Raw Data to AI/ML Insights • Minimal Learning Curve • PyArrow Integration • Pandas/Polars DataFrame Support • DuckDB SQL Support • AI/ML Ecosystem Compatibility