[
https://issues.apache.org/jira/browse/SPARK-24615?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16549266#comment-16549266
]
Thomas Graves commented on SPARK-24615:
---------------------------------------
I think the usage for cpu/memory is the same. You know one job or stage has x
number of tasks and perhaps they are caching at the time and need more memory
or cpu. For instance look at your accelerator example. Let say I'm doing some
etl before my ML. Once I get to the point I want to do the ML I want to ask
for the gpu's as well as ask for more memory during that stage because I didn't
need more before this stage for all the etl work. I realize you already have
executors, but ideally spark with the cluster manager could potentially release
the existing ones and ask for new ones with those requirements. That is
obviously a separate task to do the latter part but I think if we are creating
an interface we should keep those cases in mind.
> Accelerator-aware task scheduling for Spark
> -------------------------------------------
>
> Key: SPARK-24615
> URL: https://issues.apache.org/jira/browse/SPARK-24615
> Project: Spark
> Issue Type: Improvement
> Components: Spark Core
> Affects Versions: 2.4.0
> Reporter: Saisai Shao
> Assignee: Saisai Shao
> Priority: Major
> Labels: Hydrogen, SPIP
>
> In the machine learning area, accelerator card (GPU, FPGA, TPU) is
> predominant compared to CPUs. To make the current Spark architecture to work
> with accelerator cards, Spark itself should understand the existence of
> accelerators and know how to schedule task onto the executors where
> accelerators are equipped.
> Current Spark’s scheduler schedules tasks based on the locality of the data
> plus the available of CPUs. This will introduce some problems when scheduling
> tasks with accelerators required.
> # CPU cores are usually more than accelerators on one node, using CPU cores
> to schedule accelerator required tasks will introduce the mismatch.
> # In one cluster, we always assume that CPU is equipped in each node, but
> this is not true of accelerator cards.
> # The existence of heterogeneous tasks (accelerator required or not)
> requires scheduler to schedule tasks with a smart way.
> So here propose to improve the current scheduler to support heterogeneous
> tasks (accelerator requires or not). This can be part of the work of Project
> hydrogen.
> Details is attached in google doc. It doesn't cover all the implementation
> details, just highlight the parts should be changed.
>
> CC [~yanboliang] [~merlintang]
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
