[
https://issues.apache.org/jira/browse/ARROW-6691?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17105493#comment-17105493
]
Adam Lippai commented on ARROW-6691:
------------------------------------
[~andygrove] Skipping Tokio can be reasonable, but we still need a threadpool
or limiting the number of opened files somehow. Currently it opens and read all
the files in parallel, then it does all the computation in parallel (and AFAIK
reading a lot of files concurrently won't produce optimal performance on
linux).
> [Rust] [DataFusion] Use tokio and Futures instead of spawning threads
> ---------------------------------------------------------------------
>
> Key: ARROW-6691
> URL: https://issues.apache.org/jira/browse/ARROW-6691
> Project: Apache Arrow
> Issue Type: Sub-task
> Components: Rust, Rust - DataFusion
> Reporter: Andy Grove
> Assignee: Andy Grove
> Priority: Major
> Fix For: 1.0.0
>
> Attachments: image-2019-12-07-17-54-57-862.png
>
>
> The current implementation of the physical query plan uses "thread::spawn"
> which is expensive. We should switch to using Futures, async!/await!, and
> tokio so that we are launching tasks in a thread pool instead and writing
> idiomatic Rust code with futures combinators to chain actions together.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
