Work Stealing Criteria for stealing Computation to Communication Ratio Saturated Worker Burden Replicate Popular Data Steal from the Rich Restrictions Choosing tasks to steal...

Data Streams with Queues Data Streams with Queues This feature is no longer supported.Instead people may want to look at the following options: Use normal for loops with Clie...

Resilience User code failures Closed Network Connections Hardware Failures Scheduler Failure Restart and Nanny Processes Resilience Software fails, Hardware fails, network...

API reference Top-level functions Dataset Creating a dataset Attributes Dictionary interface Dataset contents Comparisons Indexing Missing value handling Computation Resha...

Development roadmap Our philosophy Specific needs Technical vision Flexible indexes Flexible arrays Flexible storage Engaging more users Administrative Current core develope...

操作指南 操作指南 译者：@ImPerat0R_ 在快速开始 部分设置沙箱很容易; 建设一个生产级环境则需要更多的工作！ 这些操作指南将指导您完成使用和配置Airflow环境的常见任务。 设置配置选项 初始化数据库后端 使用操作器 BashOperator PythonOperator Google云端平台运营商 管理...

Understanding Performance Understanding Performance The first step in making computations run quickly is to understand the costs involved.In Python we often rely on tools liket...

Bag: Parallel Lists for semi-structured data Create data Setup Creation Manipulation Example: Accounts JSON data Basic Queries Use flatten to de-nest Groupby and Foldby grou...

Executor Executor Types Executor Executors are the mechanism by which task instances get run. They have a common API and are “pluggable”, meaning you can swap executors based ...

Scheduling Policies Choosing Workers Choosing Tasks Last in, first out Break ties with children and depth Initial Task Placement First-Come-First-Served, Coarsely Where these ...