Скачать или смотреть OpenXData - Open Source Query Performance: Inside the next-gen Presto C++ engine

Presto (https://prestodb.io/) is a popular open source SQL query engine for high performance analytics in the Open Data Lakehouse. Originally developed at Meta, Presto has been adopted by some of the largest data-driven companies in the world including Uber, ByteDance, Alibaba and Bolt. Today it’s available to run on your own or through managed services such as IBM watsonx.data and AWS Athena.

Presto is fast, reliable and efficient at scale. The latest innovation in Presto is a state of the art C++ native query execution engine that replaces the old Java execution engine. Presto C++ is built using Velox, which is another Meta open source project for common runtime primitives across query engines. Deployments with the new Presto native engine show massive price performance improvements with fleet sizes shrinking to almost 1/3rd of their Java cluster counterparts, leading to enormous cost savings.

The Presto Native engine project began in 2020 and since then, it has matured into production use at Meta, Uber, and in IBM watsonx.data. This talk gives an in-depth look at this journey, covering:
Introduction to Prestissimo/Velox architecture
Production experiences and learnings from Meta
Benchmarking results from TPC-DS workloads
New lakehouse capabilities enabled by the native engine

Beyond the product features, Aditi Pandit from IBM and Amit Dutta of Meta highlight how the open source community shaped this innovation and the benefits of building technology like this openly across many companies.