Reliable Performance at Scale with Apache Spark on Kubernetes - Will Manning, Matthew Cheah Palantir
Description
Kubernetes is an open-source containerization framework that makes it easy to manage applications in isolated environments at scale. In Apache Spark 2.3, Spark introduced support for native integration with Kubernetes. Palantir has been deeply involved with the development of Spark's Kubernetes integration from the beginning, and our largest production deployment now runs an average of ~5 million Spark pods per day, as part of tens of thousands of Spark applications. Over the course of our adventures in migrating deployments from YARN to Kubernetes, we have overcome a number of performance, cost, & reliability hurdles: differences in shuffle performance due to smaller filesystem caches in containers; Kubernetes CPU limits causing inadvertent throttling of containers that run many Java threads; and lack of support for dynamic allocation leading to resource wastage. We intend to briefly describe our story of developing & deploying Spark-on-Kubernetes, as well as lessons learned from deploying containerized Spark applications in production. We will also describe our recently open-sourced extension (https://github.com/palantir/k8s-spark-scheduler) to the Kubernetes scheduler to better …
Description from YouTube. Full content on the video page.
More from Databricks
NewsHow adidas Uses Databricks to Build Better Products
Adidas uses Databricks' lakehouse platform to centralize all its data—from product to football-related insights—enabling faster analytics across the organization. The company's Genie analytics tool helps analysts spend less time processing data and more time on strategic questions, ultimately supporting better product development.
NewsDatabricks for Good x MapAid: Creating a Searchable Database for Groundwater Discovery
MapAid is using Databricks to build an AI groundwater mapping system called Well Mapper that processes water documents to identify accurate well-drilling locations in Ethiopia, improving on the current 30% success rate. The system analyzes 400 well logs in half a second instead of four weeks, with potential to double Ethiopia's food supply through improved irrigation.
NewsHow Databricks Genie Automates Data Workflows with Genie Ontology and Scheduled Tasks
Databricks Genie enables ontology by default for business context and adds document/PDF uploads, direct Unity Catalog queries, and team collaboration features in chat. Scheduled tasks automate recurring workflows with embedded visualizations and PDF outputs accessible across web, desktop, and mobile platforms.
NewsGenie One Beginner's Guide: Explore Data & Automate Tasks
Genie One helps subject matter experts avoid repetitive questions by using business ontology to understand your data and run real-time queries that generate instant answers and shareable reports. The tool can also automate monitoring with conditional alerts that only notify you when specific business thresholds are met.
TutorialsHow to Schedule Automated Meeting Prep in Genie One
Genie One allows users to connect multiple data sources like Databricks tables and Google Calendar, then create analyses through natural language prompts to prepare for customer meetings. The platform can schedule these analyses to run automatically on a recurring basis and deliver results via email or mobile app.
TutorialsHow to Build Custom Skills in Genie One in Minutes
Genie 1 can populate business report templates by analyzing their format and running SQL queries against organizational data. Users can save these report workflows as reusable skills that execute with a single slash command, enabling automation of recurring reports.