[
https://issues.apache.org/jira/browse/IGNITE-26724?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Kirill Tkalenko resolved IGNITE-26724.
--------------------------------------
Resolution: Duplicate
> Consider using RandomAccessFileIoFactory instead of AsyncFileIoFactory
> ----------------------------------------------------------------------
>
> Key: IGNITE-26724
> URL: https://issues.apache.org/jira/browse/IGNITE-26724
> Project: Ignite
> Issue Type: Improvement
> Reporter: Kirill Tkalenko
> Assignee: Kirill Tkalenko
> Priority: Major
> Labels: ignite-3
>
> While investigating checkpoint slowdowns in aipersist, I tried replacing
> *org.apache.ignite.internal.fileio.AsyncFileIoFactory* with
> *org.apache.ignite.internal.fileio.RandomAccessFileIoFactory* and noticed a
> performance boost for writing dirty pages and fsync files, but for a small
> number of files and pages. We need to investigate this in more detail and try
> it with a larger number of pages.
> It's also worth noting that *AsyncFileIoFactory* was introduced in AI 2.0 for
> a reason, but to address the issue of thread interrupts that were causing
> *FileChannel* to close. We need to check if this is the issue here. If the
> problem persists, we can consider using *RandomAccessFileIoFactory* for page
> writes to speed them up.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
