Self integrity performance

Question

Hi all,
(11.2.0.2 on RHEL 5.3, non RAC core OLTP db)
I am involved into a data modeling practice and i got into a situation to evaluate performance of self integrity (Multiple FKs points to PK column within same table).
I have below table structure..
  CREATE TABLE tab1
   (    ID NUMBER(32) NOT NULL ENABLE,
        Name varchar2(50),
        ref_id (32),
        tid NUMBER(32),
        CONSTRAINT pk_id PRIMARY KEY (ID)
        CONSTRAINT fk_refid FOREIGN KEY (ref_id) REFERENCES tab1 (ID) ENABLE,
        CONSTRAINT fk_tid FOREIGN KEY (tid) REFERENCES  tab1 (ID) ENABLE
     )I am not sure what could be alternate approach for such modeling. But i believe such 'self integrity model' causes IO (IOPS) issues from scalability view point when table has high growth rate with lots of updates hitting it.
Can anyone share experience here ?

Answer 1

Justin Cave wrote:
First, a data modeling question that doesn't explain what you are modeling (and thus why two self-referential foreign keys are potentially appropriate) is extremely hard to provide any feedback on. Without knowing what you are modeling, it is impossible to suggest whether you are modeling it properly or whether there are alternate ways to model that data.
I think it's also worth pointing out that the construction of the test could make a big difference to how the database performs and, in the absence of any information, no-one is in a position to have any ideas about why your perception is different from the reality seen in the production system mentioned by another user.
Whatever else your test does, you seem to be using very small rows, which means a lot of rows per block, which could mean a lot of contention and read-consistency overhead when you try to run highly concurrent inserts.
Regards
Jonathan Lewis