databendlabs / databend

𝗗𝗮𝘁𝗮, 𝗔𝗻𝗮𝗹𝘆𝘁𝗶𝗰𝘀 & 𝗔𝗜. Modern alternative to Snowflake. Cost-effective and simple for massive-scale analytics. https://databend.com
https://docs.databend.com
Other
7.9k stars 752 forks source link

Databend Task Frameworks #8020

Open Xuanwo opened 2 years ago

Xuanwo commented 2 years ago

Summary

Databend needs task frameworks so that it can:

ZhiHanZ commented 2 years ago

I think besides those features, another feature value most is the ability to handle tasks in short-lived warehouse running instances, and we may need:

  1. sync and manage task session to get avoid of session leakage. (/query/status response should have dedicates struct represents started/running/finished jobs and tasks)
  2. ability to drain task in a single instance and distribute the remaining tasks into the rest of active instances (fault tolerance)

cc @flaneur2020 any other suggestions?

b41sh commented 1 year ago

@Xuanwo Is there any progress on this issue? It is time-consuming to build BlockFilter and BlockStatistics when inserting a large number of values. If we can use an asynchronous task framework to do this can speed up the insertion operation.

Xuanwo commented 1 year ago

Is there any progress on this issue?

No progress so far.