[
https://issues.apache.org/jira/browse/CASSANDRA-1470?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12969953#action_12969953
]
Pavel Yaskevich commented on CASSANDRA-1470:
--------------------------------------------
This is more complex that it seems to be from the first sight: All Direct I/O
operations will be synchronous; operations will not return until completed. On
the other hand standard I/O (using page cache) uses mechanism called Page
Writeback - write syscalls place pages into memory (if necessary), mark them as
dirty and return immediately after that (pdflush kernel threads are responsible
for syncing page cache data to the device) actual memory<->device
synchronization happens when:
1) When free memory shrinks below a configurable threshold, dirty buffers are
written back to disk so that the now-clean buffers may be removed, freeing
memory.
2) When a dirty buffer ages beyond a configurable threshold, the buffer is
written back to disk. This prevents data from remaining dirty indefinitely.
Deferred writes and buffer subsystem in Linux enable fast writes...
More on this could be found in the book "Linux System Programming" in section
File I/O -> Kernel Internals or in the book "Understanding The Linux Kernel 3rd
Edition"...
> use direct io for compaction
> ----------------------------
>
> Key: CASSANDRA-1470
> URL: https://issues.apache.org/jira/browse/CASSANDRA-1470
> Project: Cassandra
> Issue Type: Improvement
> Components: Core
> Reporter: Jonathan Ellis
> Assignee: Pavel Yaskevich
> Fix For: 0.7.1
>
> Attachments: 1470-v2.txt, 1470.txt, CASSANDRA-1470-for-0.6.patch,
> CASSANDRA-1470-v10-for-0.7.patch, CASSANDRA-1470-v11-for-0.7.patch,
> CASSANDRA-1470-v12-0.7.patch, CASSANDRA-1470-v2.patch,
> CASSANDRA-1470-v3-0.7-with-LastErrorException-support.patch,
> CASSANDRA-1470-v4-for-0.7.patch, CASSANDRA-1470-v5-for-0.7.patch,
> CASSANDRA-1470-v6-for-0.7.patch, CASSANDRA-1470-v7-for-0.7.patch,
> CASSANDRA-1470-v8-for-0.7.patch, CASSANDRA-1470-v9-for-0.7.patch,
> CASSANDRA-1470.patch,
> use.DirectIORandomAccessFile.for.commitlog.against.1022235.patch
>
>
> When compaction scans through a group of sstables, it forces the data in the
> os buffer cache being used for hot reads, which can have a dramatic negative
> effect on performance.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
