slow mysql count because of subselect - mysql

Question

how to make this select statement more faster?
the first left join with the subselect is making it slower...
mysql> SELECT COUNT(DISTINCT w1.id) AS AMOUNT FROM tblWerbemittel w1
JOIN tblVorgang v1 ON w1.object_group = v1.werbemittel_id
INNER JOIN ( SELECT wmax.object_group, MAX( wmax.object_revision ) wmaxobjrev FROM tblWerbemittel wmax GROUP BY wmax.object_group ) AS wmaxselect ON w1.object_group = wmaxselect.object_group AND w1.object_revision = wmaxselect.wmaxobjrev
LEFT JOIN ( SELECT vmax.object_group, MAX( vmax.object_revision ) vmaxobjrev FROM tblVorgang vmax GROUP BY vmax.object_group ) AS vmaxselect ON v1.object_group = vmaxselect.object_group AND v1.object_revision = vmaxselect.vmaxobjrev
LEFT JOIN tblWerbemittel_has_tblAngebot wha ON wha.werbemittel_id = w1.object_group
LEFT JOIN tblAngebot ta ON ta.id = wha.angebot_id
LEFT JOIN tblLieferanten tl ON tl.id = ta.lieferant_id AND wha.zuschlag = (SELECT MAX(zuschlag) FROM tblWerbemittel_has_tblAngebot WHERE werbemittel_id = w1.object_group)
WHERE w1.flags =0 AND v1.flags=0;
+--------+
| AMOUNT |
+--------+
| 1982 |
+--------+
1 row in set (1.30 sec)
Some indexes has been already set and as EXPLAIN shows they were used.
+----+--------------------+-------------------------------+--------+----------------------------------------+----------------------+---------+-----------------------------------------------+------+----------------------------------------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+--------------------+-------------------------------+--------+----------------------------------------+----------------------+---------+-----------------------------------------------+------+----------------------------------------------+
| 1 | PRIMARY | <derived2> | ALL | NULL | NULL | NULL | NULL | 2072 | |
| 1 | PRIMARY | v1 | ref | werbemittel_group,werbemittel_id_index | werbemittel_group | 4 | wmaxselect.object_group | 2 | Using where |
| 1 | PRIMARY | <derived3> | ALL | NULL | NULL | NULL | NULL | 3376 | |
| 1 | PRIMARY | w1 | eq_ref | object_revision,or_og_index | object_revision | 8 | wmaxselect.wmaxobjrev,wmaxselect.object_group | 1 | Using where |
| 1 | PRIMARY | wha | ref | PRIMARY,werbemittel_id_index | werbemittel_id_index | 4 | dpd.w1.object_group | 1 | |
| 1 | PRIMARY | ta | eq_ref | PRIMARY | PRIMARY | 4 | dpd.wha.angebot_id | 1 | |
| 1 | PRIMARY | tl | eq_ref | PRIMARY | PRIMARY | 4 | dpd.ta.lieferant_id | 1 | Using index |
| 4 | DEPENDENT SUBQUERY | tblWerbemittel_has_tblAngebot | ref | PRIMARY,werbemittel_id_index | werbemittel_id_index | 4 | dpd.w1.object_group | 1 | |
| 3 | DERIVED | vmax | index | NULL | object_revision_uq | 8 | NULL | 4668 | Using index; Using temporary; Using filesort |
| 2 | DERIVED | wmax | range | NULL | or_og_index | 4 | NULL | 2168 | Using index for group-by |
+----+--------------------+-------------------------------+--------+----------------------------------------+----------------------+---------+-----------------------------------------------+------+----------------------------------------------+
10 rows in set (0.01 sec)
The main problem while the statement above takes about 2 seconds seems to be the subselect where no index can be used.
How to write the statement even more faster?
Thanks for help. MT

Answer 1

Do you have the following indexes?
for tblWerbemittel - object_group, object_revision
for tblVorgang - object_group, object_revision
for tblWerbemittel_has_tblAngebot - werbemittel_id, zuschlag
Let me know if that helps, there are a few more that I can see might help but try those first.
EDIT
Can you try these two queries and see if they run fast?
SELECT w1.id AS AMOUNT
FROM tblWerbemittel w1 INNER JOIN
(SELECT wmax.object_group,
MAX( wmax.object_revision ) AS wmaxobjrev
FROM tblWerbemittel AS wmax
GROUP BY wmax.object_group ) AS wmaxselect ON w1.object_group = wmaxselect.object_group AND
w1.object_revision = wmaxselect.wmaxobjrev
WHERE w1.flags = 0
SELECT v1.werbemittel_id
FROM tblVorgang v1 LEFT JOIN
(SELECT vmax.object_group,
MAX( vmax.object_revision ) AS vmaxobjrev
FROM tblVorgang AS vmax
GROUP BY vmax.object_group ) AS vmaxselect ON v1.object_group = vmaxselect.object_group AND
v1.object_revision = vmaxselect.vmaxobjrev LEFT JOIN
WHERE v1.flags = 0

Answer 2

While I consider I don't have sufficient data to provide a 100% correct answer, but I can throw in a handful of tips.
Forst of all, MYSQL is stupid. Bear that in mind and always rearrange your queries so that the most data is excluded at the beginning. For instance, if the last join reduced the number of results from 10k to 2k while the others don't, try swapping their positions so that each subsequent join operates on the smallest subset of data possible.
Same applies to the WHERE clause.
Also, joins tend to be slower than subqueries. I don't know if that's a rule or just something that I'm observing in my case, but you can always try to substitute a join or two with a subquery.
While I suppose this doesn't really answer your question, I hope it at least gives you an idea about where to start looking for optimisations.