[ 
https://issues.apache.org/jira/browse/CASSANDRA-21088?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18046828#comment-18046828
 ] 

Dmitry Konstantinov commented on CASSANDRA-21088:
-------------------------------------------------

e2e test results (same test as here: 
https://issues.apache.org/jira/browse/CASSANDRA-20226):  
[^CASSANDRA-21088.html] 
Profiles before and after are attached as well.
||Before||After||
|{code:java} op rate                    : 192,020 op/s [insert: 192,020 op/s]
partition rate             : 192,020 pk/s [insert: 192,020 pk/s]
row rate                   : 1,920,200 row/s [insert: 1,920,200 row/s]
latency mean               : 1.5 ms [insert: 1.5 ms]
latency median             : 1.1 ms [insert: 1.1 ms]
latency 95th percentile    : 2.6 ms [insert: 2.6 ms]
latency 99th percentile    : 10.0 ms [insert: 10.0 ms]
latency 99.9th percentile  : 29.1 ms [insert: 29.1 ms]
latency max                : 438.6 ms [insert: 438.6 ms]
total gc count             : 40
total gc memory            : 384.516 GiB
total gc time              : 7.5 seconds
avg gc time                : 188.2 ms
stddev gc time             : 56.3 ms
Total operation time       : 00:01:18 {code}|{code:java}
op rate                    : 193,566 op/s [insert: 193,566 op/s]
partition rate             : 193,566 pk/s [insert: 193,566 pk/s]
row rate                   : 1,935,661 row/s [insert: 1,935,661 row/s]
latency mean               : 1.5 ms [insert: 1.5 ms]
latency median             : 1.0 ms [insert: 1.0 ms]
latency 95th percentile    : 2.5 ms [insert: 2.5 ms]
latency 99th percentile    : 9.9 ms [insert: 9.9 ms]
latency 99.9th percentile  : 28.3 ms [insert: 28.3 ms]
latency max                : 1179.6 ms [insert: 1,179.6 ms]
total gc count             : 36
total gc memory            : 374.921 GiB
total gc time              : 7.1 seconds
avg gc time                : 196.6 ms
stddev gc time             : 29.9 ms
Total operation time       : 00:01:17 {code}|

> Minor perf optimizations around memtable put logic
> --------------------------------------------------
>
>                 Key: CASSANDRA-21088
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-21088
>             Project: Apache Cassandra
>          Issue Type: Improvement
>          Components: Local/Memtable
>            Reporter: Dmitry Konstantinov
>            Assignee: Dmitry Konstantinov
>            Priority: Normal
>             Fix For: 5.1
>
>         Attachments: CASSANDRA-21088.html, memtable_after_v2_alloc.html, 
> memtable_after_v2_cpu.html, memtable_before_alloc.html, 
> memtable_before_cpu.html
>
>          Time Spent: 20m
>  Remaining Estimate: 0h
>
> * Avoid BTree iterators allocation in ColumnsCollector.update (0.3% of 
> allocations)
>  * reduce amount of virtual calls within Cell.dataSize (Cell.dataSize costs 
> 1.62 % of CPU)
>  * avoid capturing lambda allocation in BTreeRow.clone (0.59% of allocations)
>  * Do not recalculate minLocalDeletionTime when BTreeRow is cloned (0.37 % of 
> CPU)
> allocation/CPU are taken from profile data for batch write test ( 
> CASSANDRA-20226 )



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to