You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@cassandra.apache.org by "Sam Lightfoot (Jira)" <ji...@apache.org> on 2025/12/22 16:49:00 UTC
[jira] [Commented] (CASSANDRA-19987) Direct IO support for compaction reads
[ https://issues.apache.org/jira/browse/CASSANDRA-19987?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18047081#comment-18047081 ]
Sam Lightfoot commented on CASSANDRA-19987:
-------------------------------------------
h2. Direct I/O vs Buffered I/O Read Latency During Compaction
*Test Conditions:*
* 10k read requests/s against a 1GB dataset
* Unthrottled compaction of two 80GB SSTables
* 4KB block device read-ahead
* Page cache filled prior to latency capture
||Metric||Buffered||Direct I/O||Improvement||
|*Mean*|0.45ms|0.33ms|*27% lower*|
|*Std Dev*|3.91|1.48|*62% lower*|
|*p50*|0.22ms|0.21ms|5%|
|*p90*|0.42ms|0.40ms|5%|
|*p95*|0.53ms|0.50ms|6%|
|*p99*|1.59ms|0.74ms|*53%*|
|*p99.9*|29.88ms|8.22ms|*72%*|
|*p99.99*|88.08ms|57.67ms|*35%*|
|*p99.999*|360.71ms|72.35ms|*80%*|
|*Max*|402.65ms|202.38ms|*50%*|
!image-2025-12-22-16-43-54-374.png|width=814,height=219!
Scenario:
> Direct IO support for compaction reads
> --------------------------------------
>
> Key: CASSANDRA-19987
> URL: https://issues.apache.org/jira/browse/CASSANDRA-19987
> Project: Apache Cassandra
> Issue Type: Improvement
> Components: Local/Compaction
> Reporter: Jon Haddad
> Assignee: Sam Lightfoot
> Priority: Normal
> Fix For: 5.x
>
> Attachments: image-2025-12-22-16-43-54-374.png
>
> Time Spent: 2h 20m
> Remaining Estimate: 0h
>
> If we use direct io to read SSTables during compaction, we can avoid polluting the page cache with data we're about to delete. As another side effect, we also evict pages to make room for whatever we're putting in. This unnecessary churn leads to higher CPU overhead and can cause dips in client read latency, as we're going to be evicting pages that could be used to serve those reads.
> This is most notable with STCS as the SSTables get larger, potentially evicting the entire hot dataset out of cache, but is affected by every compaction strategy.
> This is a follow up to be done after CASSANDRA-15452 since we will have an internal buffer.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@cassandra.apache.org
For additional commands, e-mail: commits-help@cassandra.apache.org