Optimize MySQL query with many subselects - mysql

Question

I have to maintain an application with SQL commands (article management).
One query looks like this:
SET #from_timestamp = '2013-07-01 00:00:00';
SET #serialnumber = '%aaaa01%';
SELECT
test1.article,
test1.serialnumber,
test2.`timestamp` as test2_timestamp,
test2.additionalinfo,
test1.`timestamp` AS test1_timestamp,
test1.text0 AS test1_text0,
test1.text1 AS test1_text1,
test1.text2 AS test1_text2,
test1.text3 AS test1_text,
test3.text1 as test3_text1,
test3.`timestamp` AS test3_timestamp,
test3.`status`,
TIMESTAMPDIFF(MINUTE, test1.`timestamp`, test3.`timestamp`) as DeltaT
FROM (
SELECT
max(`timestamp`) AS TIMESTAMP,
article,
serialnumber,
text1,
text2,
text3,
text4
FROM exampleTable
WHERE
test = 'test1' AND
`timestamp` >= #from_timestamp AND
os0 LIKE #serialnumber
GROUP BY serialnumber
) AS test1
LEFT JOIN (
SELECT
max(`timestamp`) AS TIMESTAMP,
serialnumber,
status,
text1
FROM exampleTable
WHERE
test = 'test3' AND
`timestamp` >= #from_timestamp
GROUP BY serialnumber
) AS test3
ON test1.serialnumber = test3.serialnumber
LEFT JOIN (
SELECT
max(`timestamp`) AS TIMESTAMP,
serialnumber,
article,
text1
FROM exampleTable
WHERE
LENGTH(serialnumber) = 13 AND
test = 'test2' AND
`status` = -1 AND
`timestamp` >= #from_timestamp
GROUP BY serialnumber
) AS test2
ON test1.serialnumber = test2.serialnumber
ORDER BY test1.`timestamp`
It is working with only one table where product data is saved. But as expected, this query is really slow. Is there any optimization possible?
Edit: Attached image of the explain sql result
Edit2: The table has many data (last COUNT(*) provides 39589279 ;-) )
Edit3: Attached image of the explain table result
Edit4: Please don't blame me for table definition, it is very old developed by other guys and it is not possible to change it without breaking changes.

Answer 1

First of all, you qre querying three times the same table and join afterwards. The first simplification would be to use only one select :
SELECT
article,
serialnumber,
text1,
text2,
text3,
text4,
max(case when
test = 'test1' AND
os0 LIKE #serialnumber
then `timestamp` else null end) AS test1_timestamp,
max(case when
LENGTH(serialnumber) = 13 AND
test = 'test2' AND
`status` = -1
then `timestamp` else null end) AS test2_timestamp,
max(case when
test = 'test3'
then `timestamp` else null end) AS test3_timestamp,
FROM exampleTable
WHERE
test in ('test1', 'test2', 'test3') AND
`timestamp` >= #from_timestamp
GROUP BY serialnumber
This will already give you a great improvement.
Secondly, you have to put an index on the timestamp column to improve speed.