[
https://issues.apache.org/jira/browse/IGNITE-3536?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15970742#comment-15970742
]
Alexei Kaigorodov commented on IGNITE-3536:
-------------------------------------------
In order to avoid waiting for time consuming operations, they should be
parallelised. Parallelisation can be done in 2 forms: synchronous (using
threads and both blocking and non-blocking operations) and asynchronous (using
tasks executed by thread pool and non-blocking operations only).
Asynchronous parallelisation is a hard task: blocking operations must be
eliminated totally. Program code bloats and become less readable. New
challenges like back-pressure should be resolved. The only advantage of
asynchronous parallelisation is saving memory for thread's stacks.
That is, I want to make sure we definitely want asynchronous parallelisation
and not synchronous one.
> IGFS: Implement async methods for all base file system operations.
> ------------------------------------------------------------------
>
> Key: IGNITE-3536
> URL: https://issues.apache.org/jira/browse/IGNITE-3536
> Project: Ignite
> Issue Type: Task
> Components: IGFS
> Affects Versions: 1.6
> Reporter: Vladimir Ozerov
> Assignee: Alexei Kaigorodov
> Fix For: 2.1
>
>
> 1) Remove {{IgniteAsyncSupport}} interface
> 2) Implement async counterparts for all FS operations.
> Justification: some structure file system operations might be very
> time-consuming, so having async counterparts sounds like a good idea.
> The questions is what thread pool will host these tasks.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
