[
https://issues.apache.org/jira/browse/IGNITE-20270?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Aleksandr Polovtcev updated IGNITE-20270:
-----------------------------------------
Description:
When running benchmarks that insert around 7000000 rows into Ignite 3, I've
discovered that a lot of time is spent in {{iterator.seek}} calls:
!screenshot-3.png!
I've tried multiple strategies to improve this behavior:
# Introduce Bloom Filters:
https://github.com/facebook/rocksdb/wiki/Prefix-Seek#configure-prefix-bloom-filter
# Use Tailing Iterators:
https://github.com/facebook/rocksdb/wiki/Tailing-Iterator
# Use adaptive prefix mode:
https://github.com/facebook/rocksdb/wiki/Prefix-Seek#adaptive-prefix-mode
This improved some scenarios up to 6 times:
!screenshot-2.png!
> Improve RocksDB iterator performance
> ------------------------------------
>
> Key: IGNITE-20270
> URL: https://issues.apache.org/jira/browse/IGNITE-20270
> Project: Ignite
> Issue Type: Improvement
> Reporter: Aleksandr Polovtcev
> Assignee: Aleksandr Polovtcev
> Priority: Major
> Labels: ignite-3
> Attachments: screenshot-2.png, screenshot-3.png
>
>
> When running benchmarks that insert around 7000000 rows into Ignite 3, I've
> discovered that a lot of time is spent in {{iterator.seek}} calls:
> !screenshot-3.png!
> I've tried multiple strategies to improve this behavior:
> # Introduce Bloom Filters:
> https://github.com/facebook/rocksdb/wiki/Prefix-Seek#configure-prefix-bloom-filter
> # Use Tailing Iterators:
> https://github.com/facebook/rocksdb/wiki/Tailing-Iterator
> # Use adaptive prefix mode:
> https://github.com/facebook/rocksdb/wiki/Prefix-Seek#adaptive-prefix-mode
> This improved some scenarios up to 6 times:
> !screenshot-2.png!
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
