Forum Stats

  • 3,741,441 Users
  • 2,248,430 Discussions
  • 7,861,808 Comments

Discussions

Table partitioning

simon87
simon87 Member Posts: 55 Blue Ribbon
edited Jan 6, 2015 3:28AM in SQL & PL/SQL

I currently have some tables with data and need to partition. Can be a partition to an existing table with lots of data?

How do I manipulate a partition? For example: In the query, how would he know on which partition find?, in which partition delete? What partition insert?


Thanks

simon87Cherif bh

Answers

  • L-MachineGun
    L-MachineGun Member Posts: 926

    Learn about partitioning in the fine Oracle® Database VLDB and Partitioning Guide

    To convert non-partitioned table to partitioned table, use DBMS_REDEFINITION

    simon87
  • simon87
    simon87 Member Posts: 55 Blue Ribbon

    Ok, thank you. Greetings

  • Keen2Learn
    Keen2Learn Member Posts: 103

    For creating partitions, it is better to create partitioned tables first and the copy data from non-partitioned table to partitioned table.

    About search from query, it depends on what kind of partitioning you are using and while searching, partition key should be referenced to improve query performance.

    You can select from specific partition by below:

    select * from tablename partition (partitioname);

    Similarly, you can delete from specific partition by mentioning partition clause.

    Insert is based on what kind of partitioning you using. Based on insert values, it will insert into specific partition for which data is true.

    Thank you.

    simon87
  • Billy Verreynne
    Billy Verreynne Software Engineer Member Posts: 28,470 Red Diamond
    Keen2Learn wrote:
    
    You can select from specific partition by below:
    
    select * from tablename partition (partitioname);
    
    Similarly, you can delete from specific partition by mentioning partition clause.
    
    

    Not a sound approach as it requires the SQL statement to know the partitioning scheme of the table, and the name of the partition. It also forces a literal (the partition name) to be used in the SQL, which requires a brand new SQL statement for the same query on a different partition.

    In app developer terms, it means having to use dynamic SQL to string the relevant partition name into the SQL. It also means runtime bugs as if there is data for that query and you point it to the wrong partition in your code, no rows will be found.

    Partitioning should be TRANSPARENT to the developer and database user. They should not need to know whether that table is partitioned, how it is partitioned, and what the name of the partition of interest is.

    They should write STANDARD SQL, with bind variables, and the CBO should be able to determine the relevant partition or partitions to query, the vast majority of the time.

    simon87
This discussion has been closed.