[
https://issues.apache.org/jira/browse/IGNITE-8591?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Vladimir Ozerov updated IGNITE-8591:
------------------------------------
Description:
When index page match condition, we eagerly read all matched data rows. This
leads to a number of random disk reads.as Ignite use heap-organized storage. We
can pre-sort all matched row links in accordance to their physical location,
and then read them in batch. This will give us two important advantages:
1) Data reads will be more sequential, this is especially important for HDDs
2) This could decrease number of page reads in case of dense data placement,
because there will be less evictions.
In future we should expand this optimization to several index pages in the same
way it is done in major databases.
> SQL: Sort links on index pages in physical page order before row access
> -----------------------------------------------------------------------
>
> Key: IGNITE-8591
> URL: https://issues.apache.org/jira/browse/IGNITE-8591
> Project: Ignite
> Issue Type: Task
> Components: sql
> Affects Versions: 2.5
> Reporter: Vladimir Ozerov
> Assignee: Vladimir Ozerov
> Priority: Major
> Fix For: 2.6
>
>
> When index page match condition, we eagerly read all matched data rows. This
> leads to a number of random disk reads.as Ignite use heap-organized storage.
> We can pre-sort all matched row links in accordance to their physical
> location, and then read them in batch. This will give us two important
> advantages:
> 1) Data reads will be more sequential, this is especially important for HDDs
> 2) This could decrease number of page reads in case of dense data placement,
> because there will be less evictions.
> In future we should expand this optimization to several index pages in the
> same way it is done in major databases.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
