Page MenuHomePhabricator

Co-locate Presto with Hadoop worker nodes
Closed, DeclinedPublic

Description

We never opened a task to track the work needed to co-locate Presto on Hadoop worker nodes. This will likely involve (high level):

  • establishing what part of RAM/cores will be dedicated to Presto for different hw types, and reduce Yarn's settings accordingly
  • add an-presto100x to the analytics hadoop worker nodes
  • deploy presto on all worker nodes.

Also, when do we want to have this done?