[
https://issues.apache.org/jira/browse/CASSANDRA-18216?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18114075#comment-18114075
]
Yash Sunil Nasery commented on CASSANDRA-18216:
-----------------------------------------------
After addressing the iterator comments, the flush performance for sharded is
almost the same or even 1-2% better than the unsharded index. Here are some JMH
numbers
{code:java}
[java]
[java] Benchmark (numberOfTerms) (shardCount)
Mode Cnt Score Error Units
[java] MemtableIndexFlushBench.flushBench 50 1
avgt 5 105.858 ± 2.601 ms/op
[java] MemtableIndexFlushBench.flushBench 50 4
avgt 5 104.469 ± 1.700 ms/op
[java] MemtableIndexFlushBench.flushBench 50 8
avgt 5 104.224 ± 3.258 ms/op
[java] MemtableIndexFlushBench.flushBench 100 1
avgt 5 104.977 ± 2.294 ms/op
[java] MemtableIndexFlushBench.flushBench 100 4
avgt 5 105.066 ± 3.001 ms/op
[java] MemtableIndexFlushBench.flushBench 100 8
avgt 5 104.987 ± 1.246 ms/op {code}
> Allow sharding of the SAI in-memory index
> ------------------------------------------
>
> Key: CASSANDRA-18216
> URL: https://issues.apache.org/jira/browse/CASSANDRA-18216
> Project: Apache Cassandra
> Issue Type: Improvement
> Components: Feature/SAI
> Reporter: Mike Adamson
> Assignee: Yash Sunil Nasery
> Priority: Normal
> Labels: SAI
> Fix For: 7.x
>
> Attachments: ci_summary.html, result_details.tar.gz
>
> Time Spent: 6h 20m
> Remaining Estimate: 0h
>
> The Memtable implementation allows it to be split into a number of shards.
> This reduces contention during writes.
> The MemtableIndex in SAI should follow this practice. It currently does not
> support sharding so all writes hit the same synchronized write block. The
> in-memory index search can also use the sharding to reduce the number of
> indexes that it searches based on the key range passed to the search.
> This exists already in the DS fork:
> https://github.com/datastax/cassandra/pull/298
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
