[ https://issues.apache.org/jira/browse/SPARK-24615?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16546580#comment-16546580 ]
Thomas Graves commented on SPARK-24615: --------------------------------------- The user is responsible for asking yarn for the right executors? I assume perhaps based of the node label or something like that? Yarn in like 3.x has the ability to support other resource types like gpus, it would be much more efficient if you can ask the resource manager for the necessary resources, is there plan to support that? I also don't think its very good behavior to have the job just hang if user doesn't have the necessary resources. it feels like that should either be fail or if it can be linked to dynamic allocation that would be better. I left some questions in the design doc. > 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