[
https://issues.apache.org/jira/browse/CASSANDRA-21160?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18093154#comment-18093154
]
Abhijeet Dubey edited comment on CASSANDRA-21160 at 7/2/26 7:01 AM:
--------------------------------------------------------------------
Compared write throughput with and without the patch using
[https://github.com/apache/cassandra-easy-stress/pull/86]
Setup:
# Single C* node running locally
# Baseline: `26d6b4f665` on cassandra-6.0 branch
# Patched: [https://github.com/apache/cassandra/pull/4911]
# 384 dimensional random vectors
# Only writes (no reads / deletes)
# Cosine similarity, SAI index, RF=1 (SimpleStrategy), STCS, JDK21
||target w/s||Baseline achieved w/s||Baseline write p99 (ms)||Patched achieved
w/s||Patched write p99 (ms)||
|200|110|3.1|110|2.7|
|400|220|3.5|220|2.4|
|800|424|~3268 (latency-collapsed, 0 dropped)|440|2.6|
|1600|453|~11866 (28% dropped)|881|2.5|
|3200|not tested — baseline already unusable at >=800|—|1764|3.2|
|6400|not tested — baseline already unusable at >=800|—|3387|15 (6% dropped)| |
dropped numbers are client side (will confirm by probing server during the next
run possibly via tpstats or the profiler
was (Author: JIRAUSER306221):
Compared write throughput with and without the patch using
[https://github.com/apache/cassandra-easy-stress/pull/86]
Setup:
# Single C* node running locally
# Baseline: `26d6b4f665` on cassandra-6.0 branch
# Patched: [https://github.com/apache/cassandra/pull/4911]
# 384 dimensional random vectors
# Only writes (no reads / deletes)
# Cosine similarity, SAI index, RF=1 (SimpleStrategy), STCS, JDK21
||target w/s||Baseline achieved w/s||Baseline write p99 (ms)||Patched achieved
w/s||Patched write p99 (ms)||
|200|110|3.1|110|2.7|
|400|220|3.5|220|2.4|
|800|424|~3268 (latency-collapsed, 0 dropped)|440|2.6|
|1600|453|~11866 (28% dropped, client-side)|881|2.5|
|3200|not tested — baseline already unusable at >=800|—|1764|3.2|
|6400|not tested — baseline already unusable at >=800|—|3387|15 (6% dropped,
client-side)|
dropped numbers are client side (will confirm by probing server during the next
run possibly via tpstats or the profiler
> Make synchronization on VectorMemoryIndex inserts more granular
> ---------------------------------------------------------------
>
> Key: CASSANDRA-21160
> URL: https://issues.apache.org/jira/browse/CASSANDRA-21160
> Project: Apache Cassandra
> Issue Type: Improvement
> Components: Feature/Vector Search
> Reporter: Michael Marshall
> Assignee: Caleb Rackliffe
> Priority: Normal
> Fix For: 5.0.x, 6.x, 7.x
>
> Attachments: ci_summary-1.html, ci_summary-2.html, ci_summary.html,
> result_details.tar-1.gz, result_details.tar-2.gz, result_details.tar.gz
>
> Time Spent: 1h 20m
> Remaining Estimate: 0h
>
> The `VectorMemoryIndex` supports asynchronous additions to the graph.
> However, the `synchronous` keyword for the `add` method prevents this and
> therefore means we are adding a point of contention that is unnecessary.
> As part of this ticket, we should verify my understanding that writes can be
> concurrent and then drop the keyword.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
