[jira] [Created] (IGNITE-7026) Index-organized data storage format

["Vladimir Ozerov (JIRA)" <ji...@apache.org>]

Vladimir Ozerov created IGNITE-7026:
---------------------------------------

             Summary: Index-organized data storage format
                 Key: IGNITE-7026
                 URL: https://issues.apache.org/jira/browse/IGNITE-7026
             Project: Ignite
          Issue Type: Task
          Components: cache, sql
            Reporter: Vladimir Ozerov


In SQL *index-organized* table is a type of table format where rows are stored as leafs of a primary key index (sometimes called "clustered index"). In this format data within a single page is sorted in accordance with PK index. All leaves are always sorted as well. 

Another table format is *heap*. Data is put into arbitrary page with enough space. Free space is tracked using either free-lists or allocation maps. Primary key index is organized in the same way as secondary index - leaf pages contain a kind of row pointer. This is how Ignite currently works. 

This ticket is aimed to implement index-organized storage format, which will give us the following advantages:
1) Fast scans over PK index due to decreased number of page reads and page locks, which is especially important for JOINs and OLAP cases;
2) Faster inserts in OLTP workloads due to less number of page updates.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)