Optimization of Mysql Join Statement Queries - mysql

Question

SELECT GV.ID
, XS.SymbolId
, GS.ID
, XS.SymbolExchangeId
, XS.IssueId
, GE.ID
, XD.ACTIVE
, XD.ExchangeId
FROM TB_GDS_SECURITY GS
, WSOD_Vanilla.WSOD_XrefIssueSymbols XS
, WSOD_Vanilla.WSOD_XrefIssueData XD
, TB_GDS_EXCHANGE GE
, TB_GDS_VENDOR GV
WHERE XD.CompositeIssueID = GS.ISSUE_ID_TEMP
AND XD.IssueId = XS.IssueId
AND XD.ACTIVE = 'True'
AND GV.VENDOR_SHORT_NAME = XS.SymbolsetId
AND GE.EXCHANGE_SHORT_NAME = XD.ExchangeId
AND NOT EXISTS (SELECT ID
FROM TB_GDS_VENDOR_VENUE_SYMBOL VVS
WHERE VVS.ISSUE_ID = XS.IssueId
AND (SELECT ID
FROM TB_GDS_VENDOR GV
WHERE GV.VENDOR_SHORT_NAME = XS.SymbolsetId
)
);
This Particular query takes more than a minute. I want it to be done in lesser time.
the explain statement is as follow:
mysql> EXPLAIN SELECT GV.ID, XS.SymbolId, GS.ID, XS.SymbolExchangeId, XS.IssueId, GE.ID, XD.ACTIVE, XD.ExchangeId FROM TB_GDS_SECURITY GS, WSOD_Vanilla.WSOD_XrefIssueSymbols XS, WSOD_Vanilla.WSOD_XrefIssueData XD, TB_GDS_EXCHANGE GE, TB_GDS_VENDOR GV WHERE XD.CompositeIssueID = GS.ISSUE_ID_TEMP AND XD.IssueId = XS.IssueId AND XD.ACTIVE = 'True' AND GV.VENDOR_SHORT_NAME=XS.SymbolsetId AND GE.EXCHANGE_SHORT_NAME = XD.ExchangeId AND NOT exists (SELECT ID FROM TB_GDS_VENDOR_VENUE_SYMBOL VVS
-> WHERE VVS.ISSUE_ID = XS.IssueId AND (SELECT ID FROM TB_GDS_VENDOR GV WHERE GV.VENDOR_SHORT_NAME = XS.SymbolsetId));
+----+--------------------+-------+------+--------------------------------------------------------+-----------------------------+---------+----------------------------------+---------+------------------------------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+--------------------+-------+------+--------------------------------------------------------+-----------------------------+---------+----------------------------------+---------+------------------------------------+
| 1 | PRIMARY | XD | ref | WSOD_XrefIssueData_UINDX1,WSOD_XrefIssueData_INDX1 | WSOD_XrefIssueData_INDX1 | 21 | const | 3981788 | Using index condition; Using where |
| 1 | PRIMARY | GE | ref | TB_GDS_EXCHANGE_INDX1 | TB_GDS_EXCHANGE_INDX1 | 63 | WSOD_Vanilla.XD.ExchangeId | 1 | Using where; Using index |
| 1 | PRIMARY | GS | ref | TB_GDS_SECURITY_INDX1 | TB_GDS_SECURITY_INDX1 | 5 | WSOD_Vanilla.XD.CompositeIssueID | 1 | Using where; Using index |
| 1 | PRIMARY | XS | ref | PRIMARY | PRIMARY | 8 | WSOD_Vanilla.XD.IssueId | 1 | Using where |
| 1 | PRIMARY | GV | ref | TB_GDS_VENDOR_INDX1 | TB_GDS_VENDOR_INDX1 | 62 | WSOD_Vanilla.XS.SymbolsetId | 1 | Using where; Using index |
| 2 | DEPENDENT SUBQUERY | VVS | ref | TB_GDS_VVENUE_SYMBOL_UINDX1,TB_GDS_VVENUE_SYMBOL_INDX2 | TB_GDS_VVENUE_SYMBOL_UINDX1 | 5 | WSOD_Vanilla.XS.IssueId | 3 | Using where; Using index |
| 3 | DEPENDENT SUBQUERY | GV | ref | TB_GDS_VENDOR_INDX1 | TB_GDS_VENDOR_INDX1 | 62 | WSOD_Vanilla.XS.SymbolsetId | 1 | Using where; Using index |
+----+--------------------+-------+------+--------------------------------------------------------+-----------------------------+---------+----------------------------------+---------+------------------------------------+
7 rows in set (0.01 sec)
After optimizing still it takes around 1.35 min to execute.
Please help me out.

Answer 1

As per comment elsewhere, you've not shown us the index defintions.
Most likely the bog cost here is the NOT EXISTS sub-clause. MySQL does not handle push-predictates well, and you've also got a cartesian product in there. I'm struggling to understand what this actually does. MySQL (indeed no DBMS I'm familiar with) can effectively use an index for NOT EXISTS or <> or NOT LIKE. The nested nested query does not appear to serve any function since you already have an explicit join between these tables in the outer query:
WHERE...AND GV.VENDOR_SHORT_NAME = XS.SymbolsetId...
AND NOT EXISTS ( [a condition] AND
SELECT ID FROM TB_GDS_VENDOR GV
WHERE GV.VENDOR_SHORT_NAME = XS.SymbolsetId
Your only filtering apart from the joins and this ugly NOT EXISTS is "AND XD.ACTIVE = 'True'" - if this column only holds 2/3 values then it will likely be very inefficient to use an index.