Presto Foundation

⭐ Presto 0.295 is out ⭐ This release adds OAuth2 authentication, native spatial joins, cross-cluster query retry, mTLS for Arrow Flight, and new PostgreSQL GEOMETRY support, plus a few key configuration and documentation improvements. Note on some breaking changes: Presto now requires Java 17 for coordinators and workers, and inline SQL-invoked functions have moved to a new plugin (presto-sql-invoked-functions-plugin). Full release notes → https://lnkd.in/dAXBsgj2 Download → https://lnkd.in/gsPTWn_z Thanks to everyone in the community who contributed to this release!

Join us next week for our October virtual meetup! We'll have Adrian Carpente Recouso from Denodo sharing more about the work they've done on support for the MERGE command in the Presto engine and Iceberg connector, a SQL-native way to perform UPSERT operations on Iceberg tables. 📅 Thurs, Oct 23 at 9am PT 💻 Join us on LinkedIn live or Zoom (rsvp thru meetup.com!)

Velox now runs at GPU speed 🚀 IBM and NVIDIA have teamed up to bring cuDF-powered GPU execution to Velox delivering big gains for #Presto and Apache Gluten. Benchmarks: • Presto TPC-H SF1,000: up to 12× faster on GPU vs. CPU • Multi-GPU Presto on DGX A100: >6× faster with UCX-based GPU Exchange over NVLink Velox now pushes scans, joins, and aggregations to the GPU using reusable operators that accelerate workloads across the open data stack. Check out the NVIDIA blog for more details https://lnkd.in/gGxFZK3a A big thank you to the engineers who have worked to bring this to fruition! Zoltán Arnold Nagy Gregory Kimball Devavret Makkar Daniel Bauer Chengcheng Jin Deepak Majeti Luis Garcés-Erice Sean Rooney Karthikeyan Natarajan Yuan ZHOU

You may have seen our announcement that the Presto engine has officially upgraded to Java 17 👏 To dig into what this means for developers and operators, we’re hosting a virtual meetup on the JDK upgrade. Zac Blanco will cover: - Why the move to Java 17 matters - Key performance and stability improvements - Compatibility considerations and upgrade steps If you’re running Presto in production or just curious about what’s next, join us on LinkedIn live to get the details straight from the community. RSVP now, we hope to see you there!

🚀 Presto 0.294 is here! This release brings a wave of improvements across performance, flexibility, and native execution. A few highlights: 🔹 Faster queries via pushdown for map_filter() and map_subset() 🔹 Boosted performance for distinct aggregations & semi joins 🔹 Case-sensitive support for schema, table, and column names 🔹 New math functions: dot_product() and l2_squared() 🔹 Updates to Router, Presto C++, Arrow Flight, and more Release notes: https://lnkd.in/eQeEXaqe Download Presto: https://lnkd.in/gsPTWn_z Huge thanks to all the contributors who made this release possible 💙

We’re kicking off a new hands-on virtual workshop series for the Presto community! Over the next few weeks, you’ll get the chance to learn by doing with three focused sessions: 🔹 Getting Started with Presto 🔹 Building a Lakehouse with Presto + Hudi + Iceberg 🔹 Diving into and Benchmarking Presto C++ Whether you’re new to Presto or looking to go deeper with the next-gen C++ engine, these workshops are designed to get you practical experience you can bring back to your own data projects. Registration is free and open now: https://lnkd.in/d3Dufvqq Join us and roll up your sleeves with Presto 😀

📣 Presto open-source announcement: JDK upgrade 👇 We’re excited to share we've upgraded the Presto core engine to Java 17, bringing performance, stability, and security improvements across the board. What this means for our Presto users: ⭐ Faster queries with modern JVM optimizations (hello, sub-millisecond GC pauses). ⭐ Stronger security with updated TLS, encryption, and Jetty 12 support. ⭐ A modern foundation for future development with new Java language features and tooling. ⭐ Client drivers and external connectors remain Java 8+ compatible, but coordinator and worker nodes now require Java 17 to take full advantage of these upgrades. Check out the full details (including config flags and migration notes) in the blog: https://lnkd.in/gPpTeKRE This is a big milestone for the project and sets the stage for what’s ahead. Big shout out to Zac Blanco for helping to drive to this huge milestone, and all of our community who worked hard to make the upgrade happen.

Scaling AI Training Data with #Prestissimo At Meta, AI training datasets are exploding in size and complexity. To meet this challenge, the #Presto team built Prestissimo - an extension that brings normalized AI training data storage and high-performance querying to massive ML workloads. In the latest deep dive, we explore: ➜ Normalized storage to cut feature duplication and storage costs. ➜ The 𝗦𝗦𝗧 𝗳𝗶𝗹𝗲 𝗳𝗼𝗿𝗺𝗮𝘁 and Presto’s new connector for fast sequential & random reads. ➜ Index Lookup Joins with Velox for low-latency, on-demand feature retrieval. Prestissimo connects Presto to Meta’s AI data infrastructure, empowering ML engineers to explore and process training data at unprecedented scale. 📖 Read the full blog: https://lnkd.in/gR6bbT28 ▶️ Checkout the PrestoCon talk: https://lnkd.in/gUm9czsF #Presto #PrestoCon #MachineLearning #OpenSource #BigData

𝗨𝗻𝗹𝗼𝗰𝗸 𝗗𝘆𝗻𝗮𝗺𝗶𝗰 𝗟𝗼𝗮𝗱𝗶𝗻𝗴 𝗳𝗼𝗿 𝗖++ 𝗨𝗗𝗙𝘀 𝗶𝗻 #𝗣𝗿𝗲𝘀𝘁𝗼 Presto C++ now supports dynamic loading of unfenced user-defined functions (UDFs), enabling developers to extend functionality without rebuilding the entire binary. This feature enables you to: ➜ Load custom C++ functions as plugins - no full rebuilds required ➜ Understand fenced vs. unfenced UDFs and their performance/security trade-offs ➜ Easily integrate modular plugins to keep custom logic separate from the core engine If you’re looking to streamline your Presto development and improve flexibility, this article is a must-read. 📖 Read the full blog: https://lnkd.in/g2aiHiPH ▶️ Checkout the PrestoCon Talk: https://lnkd.in/g6_KYjHi #Presto #Velox #UDF #BigData #OpenSource

We’re excited to share a deep dive into the 𝗧𝗣𝗖𝗗𝗦 𝗰𝗼𝗻𝗻𝗲𝗰𝘁𝗼𝗿 in #Presto C++. This connector makes it easy to generate TPCDS benchmark data dynamically for testing and performance analysis. Here’s what makes it interesting: ➜ C++ port of 𝗱𝘀𝗱𝗴𝗲𝗻 for fast, thread-safe, parallel data generation ➜ Seamless integration with 𝗩𝗲𝗹𝗼𝘅, converting connector handles and splits for efficient query execution ➜ A clear example of building and extending connectors in Presto C++ If you’re curious about benchmarking, extending Presto, or exploring its C++ ecosystem, this is a must-read. 📖 Read the full blog: https://lnkd.in/gTpZzZrB ▶️ Checkout the PrestoCon Talk: https://lnkd.in/gNzNwTaM #Presto #Velox #PrestoCon #OpenSource #BigData