Flink window

 Tue 02 February 2021   In 大数据   :)

note

event time
    数据中附带的时间
    延迟稍高(因为数据乱序以及数据晚到)
process time
    和机器有关,某个operator收到数据的时间

window会自动对齐

window components

3 parts:

  • 分组 (为下游算子多实例并行计算提供可能性)
  • 窗口分配器(WindowAssigner)
  • 触发器 (window function: reduce,aggregate,process)
  • 清除器

可自定义,否则就使用默认的配置

window type

  • 分类 tumbling window no overlap sliding window with overlap session window punctuated by a gap of inactivity 窗口的长度是可变的,每个窗口的开始和结束时间并不是确定的。

  • time-based window TimeWIndow: [startTime, endTime) time driving

  • count-based window data driving

reference

flink stream time