[
https://issues.apache.org/jira/browse/IGNITE-23975?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17906669#comment-17906669
]
Aleksey Plekhanov commented on IGNITE-23975:
--------------------------------------------
JMH results:
Before fix:
{noformat}
Benchmark (engine) Mode Cnt Score
Error Units
JmhSqlBenchmark.querySimpleUniqueIndexed H2 thrpt 3 13844,895 ±
1691,713 ops/s
JmhSqlBenchmark.querySimpleUniqueIndexed CALCITE thrpt 3 10368,102 ±
147,313 ops/s
{noformat}
After fix:
{noformat}
Benchmark (engine) Mode Cnt Score
Error Units
JmhSqlBenchmark.querySimpleUniqueIndexed H2 thrpt 3 13817,272 ±
940,452 ops/s
JmhSqlBenchmark.querySimpleUniqueIndexed CALCITE thrpt 3 11096,458 ±
158,598 ops/s
{noformat}
> Calcite engine. Group partition reservation
> -------------------------------------------
>
> Key: IGNITE-23975
> URL: https://issues.apache.org/jira/browse/IGNITE-23975
> Project: Ignite
> Issue Type: Improvement
> Reporter: Aleksey Plekhanov
> Assignee: Aleksey Plekhanov
> Priority: Major
> Labels: calcite, ise
> Time Spent: 10m
> Remaining Estimate: 0h
>
> Partition reservation performed currently one by one. We can implement group
> reservation as already implemented for H2 engine. In current implementation
> all primary partitions are reserved if {{null}} is passed as partitions
> array. For Calcite engine we can also pass {{null}} as {{ColocationGroup}}
> assignment, when assignment equals to primary partitions assignment for
> cache.
> With this fix we can achive two goals:
> 1. Faster reservation (no need to iterate over all partitions)
> 2. Faster marshalling (no need to marshal and send partitions assignment,
> when it equals to primary partitions assignment)
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
