[
https://issues.apache.org/jira/browse/IGNITE-27147?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ivan Bessonov updated IGNITE-27147:
-----------------------------------
Component/s: storage engines ai3
> Optimise compaction of multiple delta files
> -------------------------------------------
>
> Key: IGNITE-27147
> URL: https://issues.apache.org/jira/browse/IGNITE-27147
> Project: Ignite
> Issue Type: Improvement
> Components: storage engines ai3
> Reporter: Filipp Shergalis
> Assignee: Filipp Shergalis
> Priority: Major
> Labels: ignite-3
> Fix For: 3.2
>
> Time Spent: 1h 20m
> Remaining Estimate: 0h
>
> We process delta files from the oldest to the newest, and pages can be
> written to main file multiple times.
> For each delta file we already have in memory sorted arrays with page
> indexes. We can use "two pointer technique" to skip duplicated pages in
> linear time (n - pages in all other delta files).
> In the beginning of partition compaction create an int array containing
> indexes for each delta file. Processing each page iterate over all other
> delta files until we get page >= current one and skip the page if found. Next
> page continues from the same index, so it's efficient
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
