[
https://issues.apache.org/jira/browse/IGNITE-27355?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ivan Bessonov updated IGNITE-27355:
-----------------------------------
Description:
* When you add metrics for partitions within the table (for example) you copy
the entire array {{N}} times ({{{}N{}}} - number of local partitions). This is
{{O(N*M)}} ({{{}M{}}} - current number of metrics, which is at least {{O(N)}}
itself).
* When you delete the table, you do the same thing (and it uses
{{{}java.util.Collection#removeIf{}}})
We should use a concurrent hash map instead.
was:
* When you add metrics for partitions within the table (for example) you copy
the entire array {{{{N}}}} times ({{{}{{N}}{}}} - number of local partitions).
This is {{{{O(N*M)}}}} ({{{}{{M}}{}}} - current number of metrics, which is at
least {{{{O(N)}}}} itself).
* When you delete the table, you do the same thing (and it uses
{{{}java.util.Collection#removeIf{}}})
We should use a concurrent hash map instead.
> OTLP MetricReporter is inefficient
> ----------------------------------
>
> Key: IGNITE-27355
> URL: https://issues.apache.org/jira/browse/IGNITE-27355
> Project: Ignite
> Issue Type: Improvement
> Components: metrics framework ai3
> Reporter: Ivan Bessonov
> Priority: Major
> Labels: ignite-3
>
> * When you add metrics for partitions within the table (for example) you copy
> the entire array {{N}} times ({{{}N{}}} - number of local partitions). This
> is {{O(N*M)}} ({{{}M{}}} - current number of metrics, which is at least
> {{O(N)}} itself).
> * When you delete the table, you do the same thing (and it uses
> {{{}java.util.Collection#removeIf{}}})
> We should use a concurrent hash map instead.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
