[
https://issues.apache.org/jira/browse/SPARK-29791?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
zengrui updated SPARK-29791:
----------------------------
Description: We can config the executor cores by "spark.executor.cores".
For example, if we config 8 cores for a executor, then the driver can only
scheduler 8 tasks to this executor concurrently. In fact, most cases a task
does not always occupy a core or more. More time, tasks spent on disk IO or
network IO, so we can make driver to scheduler more than 8 tasks(virtual the
cores to 16，32 or more the executor report to driver) to this executor
concurrently, it will make the whole job execute more quickly. (was: We can
config the executor cores by "spark.executor.cores". For example, if we config
8 cores for a executor, then the driver can only scheduler 8 tasks to this
executor concurrently. In fact, most cases a task does not always occupy a core
or more. More time, tasks spent on disk IO or network IO, so we can make driver
to scheduler more than 8 tasks to this executor concurrently, it will make the
whole job execute more quickly.)
> Add a spark config to allow user to use executor cores virtually.
> -----------------------------------------------------------------
>
> Key: SPARK-29791
> URL: https://issues.apache.org/jira/browse/SPARK-29791
> Project: Spark
> Issue Type: Improvement
> Components: Scheduler
> Affects Versions: 2.1.0
> Reporter: zengrui
> Priority: Minor
>
> We can config the executor cores by "spark.executor.cores". For example, if
> we config 8 cores for a executor, then the driver can only scheduler 8 tasks
> to this executor concurrently. In fact, most cases a task does not always
> occupy a core or more. More time, tasks spent on disk IO or network IO, so we
> can make driver to scheduler more than 8 tasks(virtual the cores to 16，32 or
> more the executor report to driver) to this executor concurrently, it will
> make the whole job execute more quickly.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
