Conference orarep::nomahs::dbintegrator_public_public

1064.0. "Initial query by DBI ?" by ORAREP::STKHLM::KNORN (Soon in charge of a quartet) Fri Jun 07 1996 11:05

    
    I've started to analyse RDB performance on a database, which
    is accessed via DBI (on the same node).
    
    Setting the normal RDMS$DEBUG_FLAGS = S, I've noticed that 
    DBI makes an initial RDB-request to target database, before
    the 'real' querie is executed. From what I see, DBI is doing
    a cardinality request on the first table mentioned in the real
    query.
    Mostly this results in an index only retrieval, which isn't to bad,
    but sometimes it results in a sequential scan of the table.
    In the current customer case the tables are small initally, but over
    time they'll contain a large number of records, which will make
    the initial DBI-query cost a bit.
    
    Are my assumptions correct ?
    If so, why does DBI do the initial query and is there any means of
    influencing it to get a decent optimization ?
    
    Stefan

    
    Hi Stefan,
    
    What you are seeing is the network cost query getting sent over the
    wire to the rdb database so we can find out how many seconds/byte
    of data transferred (this is used in the optimizer).
    
    The query is of the form: select * from table limit to 1 row, so it
    shouldn't cause problems even if the table is huge.  Note that this is
    the query sent to Rdb - the query changes based on the database we are
    talking to.
    
    Hope this helps,
    
    dan