Home » RDBMS Server » Performance Tuning » SQL Query with many joins
SQL Query with many joins [message #139721] Thu, 29 September 2005 05:34 Go to next message
sebastianR
Messages: 33
Registered: August 2005
Member
Hi,

I'd like to know which possibilities exist to optimize a SQL-Query where there are many INNER JOINS?

The query looks like this:
============================

SELECT count(*) FROM 
(SELECT MAINTABLE.KEY_ID AS OID,
CONCAT (CONCAT (CLASSTAB.NAME, '_'),MAINTABLE.KEY_ID) AS LONGNAME, 
MAINTABLE.KEY_ID AS SHORTNAME,
'testing' as TestVal, 
MAINTABLE_COMPO_MAIN.MATURITY_DATE AS MATURITYDATE, 
MAINTABLE.ISSUE_PRICE AS FACEVALUE,
PRICETAB.quot_val AS PRICE ,
To_Date('10.02.2005','dd.mm.yyyy')  AS ENDTIME,
MAINTABLE_COMPO_INTR.FLAT_RATE AS COUPONRATE, 
MARKET.domi_ID AS COUNTRY,
PRICETAB.mkt_ID AS MARKET,
MAINTABLE.ISSUE_DATE AS DATEOFISSU,
MAINTABLE.ISSUE_PRICE AS ISSUEPRICE, 
MAINTABLE.OPN_DATE AS DATEOFCREATION, 
MAINTABLE_COMPO_MAIN.LAST_STATUS_CHANGE_DATE AS DATEOFMODIFICATION, 
PRICETAB.price_curry_id AS CURRENCY, 
MAINTABLE_COMPO_INTR.FREQ AS COUPONINTERVAL, 
(SELECT name FROM MAPTAB WHERE CLASS_ID = MAINTABLE_compo_main.DFLT_PRICE_QTYPE_CLASS_ID) AS BondQuotModeName, 
(SELECT name FROM MAPTAB WHERE CLASS_ID = MAINTABLE_compo_intr.INTR_CALC_METHOD_CLASS_ID) AS DayCountConvention 
FROM 
MAINTABLE 
INNER JOIN CLASSINST ON MAINTABLE.KEY_ID =CLASSINST.OBJ_ID 
Inner Join PRICETAB on MAINTABLE.key_id=PRICETAB.asset_id 
INNER JOIN CLASSTAB ON CLASSINST.CLASS_ID =CLASSTAB.KEY_ID 
INNER JOIN CLASSTABIF ON CLASSTAB.CLASSIF_ID = CLASSTABIF.KEY_ID 
INNER JOIN MAINTABLE_COMPO_INTR ON MAINTABLE.KEY_ID =MAINTABLE_compo_intr.asset_id 
INNER JOIN MAINTABLE_mkt ON MAINTABLE.key_id = MAINTABLE_MKT.asset_ID 
INNER JOIN MARKET ON MAINTABLE_mkt.mkt_id = MARKET.key_id 
INNER JOIN MAINTABLE_compo_main ON MAINTABLE.key_id =MAINTABLE_COMPO_MAIN.ASSET_ID  
WHERE CLASSTABIF.INTL_ID = LOWER ('test1')
AND (CLASSTAB.INTL_ID LIKE LOWER ('arg1') 
OR CLASSTAB.INTL_ID LIKE LOWER ('arg2'))
AND MAINTABLE_MKT.IS_MOST_LIQUID = 1 
AND MAINTABLE_MKT.PRIO = 1 
AND  MAINTABLE_mkt.End_Date is Null 
and MARKET.IS_TRADE_PLACE = 1 
AND MAINTABLE_COMPO_MAIN.END_DATE IS NULL 
AND MAINTABLE_COMPO_INTR.END_DATE IS NULL  
and PRICETAB.price_class_id = 1) 
WHERE  Currency In (SELECT CURR_ID FROM TABLE1) 
and MaturityDate>=EndTime 


============================

As you can see, there are 8 (!) inner joins which have to be made. They have pretty different costs, so I guess the join order is important. Here is the respective explain plan to the above query:
============================

   
Operation Object Name Rows Bytes Cost Object Node In/Out PStart PStop

SELECT STATEMENT Optimizer Mode=CHOOSE  1     43                           
 SORT AGGREGATE  1   129                             
   HASH JOIN SEMI  1   129   43                           
     NESTED LOOPS  1   116   40                           
       NESTED LOOPS  1   106   38                           
         NESTED LOOPS  1   96   36                           
           NESTED LOOPS  1   88   35                           
             NESTED LOOPS  2   144   31                           
               NESTED LOOPS  14   784   3                           
                 MERGE JOIN CARTESIAN  14   714   3                           
                   TABLE ACCESS BY INDEX ROWID PKG.CLASSTAB 1   19   2                           
                     NESTED LOOPS  1   39   3                           
                       INDEX FULL SCAN PKG.IX_CLASSTABIF_2 1   20   1                           
                       INLIST ITERATOR                                 
                         INDEX RANGE SCAN PKG.IX_CLASSTAB_1 1     1                           
                   BUFFER SORT  1   12   1                           
                     INDEX FULL SCAN PKG.UQ_PRICE_1 1   12                             
                 INDEX UNIQUE SCAN PKG.PK_MAINTABLE 1   5                             
               TABLE ACCESS BY INDEX ROWID PKG.MARKET 1   16   2                           
                 INDEX RANGE SCAN PKG.IX_MAINTABLE_MKT_1 5     1                           
             TABLE ACCESS BY INDEX ROWID PKG.MAINTABLE_COMPO_MAIN 1   16   2                           
               INDEX RANGE SCAN PKG.IX_MAINTABLE_COMPO_MAIN_2 1     1                           
           TABLE ACCESS BY INDEX ROWID PKG.MARKET 1   8   1                           
             INDEX UNIQUE SCAN PKG.PK_MARKET 2                               
         TABLE ACCESS BY INDEX ROWID PKG.MAINTABLE_COMPO_INTR 20 K 201 K 2                           
           INDEX RANGE SCAN PKG.IX_MAINTABLE_COMPO_INTR_1 1     1                           
       INDEX RANGE SCAN PKG.IX_CLASSINST_1 1   10   2                           
     TABLE ACCESS FULL PKG.TABLE1 409   5 K 2    




This query takes between 45 seconds to 3 minutes. The most expensive joins happen between MAINTABLE - CLASSINST and CLASSINST - CLASSTAB.
At least I think they are the most expensive because the joined tables return about 2 500 000 rows. Without these two joins the whole query takes only about 15 to 20 seconds.

Is there a way to make this faster? Does it make sense to split the sql-query in more than one statement?

Every help is appreciated, thank you,

sebastian
Re: SQL Query with many joins [message #139760 is a reply to message #139721] Thu, 29 September 2005 08:50 Go to previous messageGo to next message
smartin
Messages: 1803
Registered: March 2005
Location: Jacksonville, Florida
Senior Member
Do you have statistics on all of these tables? It doesn't look like it, because every line is showing only 1 row, but I'm just taking a wild guess. And what is your version? And is that a cartesian product you are doing? Are you forgetting a join condition? Also you want the optimizer mode for at least this query to be all rows.

I would write a query like this in a totally different style, which isn't necessarily better or worse just how I "see" it. But, often if you have one or two large tables slowing things down you can put them into inline views and apply any restrictive where conditions on them inside the view, so that the result that gets combined with the outer query has far far fewer rows to join.

In general you will be worse off by separating it into multiple queries, typically if you can do it in a single sql operation you should.
Re: SQL Query with many joins [message #139797 is a reply to message #139760] Thu, 29 September 2005 11:21 Go to previous messageGo to next message
sebastianR
Messages: 33
Registered: August 2005
Member
Hi smartin!

Thank you again, for your response. First of all, I didn't write this sql-statement, it was handed to me with the task to optimize it.
I didn't knew, but it seems nobody gathered statistics for the entire scheme on the server where these queries run.
So after I gathered statistics for every respective table *boom*, the execution time has gone down to 4 seconds! It was 1:30 before!

Is this possible? This improvement seems to be too enormous to be true Smile

Thanks for pointing this out, you've helped me a lot,

Sebastian
Re: SQL Query with many joins [message #139805 is a reply to message #139721] Thu, 29 September 2005 12:06 Go to previous message
smartin
Messages: 1803
Registered: March 2005
Location: Jacksonville, Florida
Senior Member
It is all about the statistics. The whole point of the optimizer is to make your statement go as fast as it can. The only way for it to do its jobis for it to have info about the data.
Previous Topic: how to rebuild index
Next Topic: Last n records based on date
Goto Forum:
  


Current Time: Thu Apr 18 04:27:31 CDT 2024