MySQL query performance limit? - mysql

Question

We have a home grown document management system and our system is running very slow, particularly on the search. It worked fine at first, but it has gotten progressively slower over time. Its now taking anywhere from 30 to 150 seconds to return results depending upon criteria. This is our search query. We’ve been staring at this thing left and right and can’t see any place to tune this more. All of the joined fields are indexed on their respective tables.
SELECT DISTINCT f.*, ts.*, fo.*, ft.*, p.*, u.*, c.*, co.*, ct.*, fs.*, fd.*, r.*, rt.*, si.*, s.* FROM (
SELECT DISTINCT f.* FROM files f
JOIN folders fo ON(fo.id = f.belongs_to_folder_id)
JOIN projects p ON(p.id = f.belongs_to_project_id)
LEFT OUTER JOIN file_statuses fs ON(fs.id = f.file_status_id)
LEFT OUTER JOIN submittal_items_files sif ON(sif.file_id = f.id)
LEFT OUTER JOIN submittal_items si ON(si.id = sif.submittal_item_id)
LEFT OUTER JOIN submittals s ON(s.id = si.belongs_to_submittal_id)
LEFT OUTER JOIN record_types rt ON(rt.id = f.record_type_id)
LEFT OUTER JOIN companies co ON(co.id = f.company_id)
LEFT JOIN folders_actions_groups ag ON (
f.belongs_to_folder_id = ag.folder_id AND
ag.action_id = 10010
)
LEFT JOIN files_actions_groups fg ON (fg.file_id = f.id)
JOIN users_groups ug ON ((ug.group_id = ag.group_id OR ug.group_id = fg.group_id) AND ug.user_id = 411)
WHERE (
(f.file_generated_name LIKE CONCAT('%', 'the', '%')) OR
(f.record_id LIKE CONCAT('%', 'the', '%')) OR
(f.record_title LIKE CONCAT('%', 'the', '%')) OR
(f.additional_info LIKE CONCAT('%', 'the', '%')) OR
(si.item_number LIKE CONCAT('%', 'the', '%')) OR
(s.element_number LIKE CONCAT('%', 'the', '%'))
) AND f.path LIKE CONCAT('Some Text', '%') AND
f.file_status_id = 3 AND
f.file_revision = 1 AND
f.discipline_id = 1 AND
f.record_type_id = 2 AND
f.triage_status_id = 2 AND
f.deleted = 0
ORDER BY f.created DESC, f.id DESC
LIMIT 100
) AS f
LEFT OUTER JOIN users u ON(f.created_by_user_id = u.id)
LEFT OUTER JOIN contacts c ON(c.user_id = u.id)
LEFT OUTER JOIN companies co ON(co.id = f.company_id)
LEFT OUTER JOIN company_types ct ON(ct.id = co.company_type_id)
JOIN triage_statuses ts ON(f.triage_status_id = ts.id)
JOIN folders fo ON(fo.id = f.belongs_to_folder_id)
JOIN folder_types ft ON(ft.id = fo.folder_type_id)
JOIN projects p ON(p.id = f.belongs_to_project_id)
LEFT OUTER JOIN file_statuses fs ON(fs.id = f.file_status_id)
LEFT OUTER JOIN file_disciplines fd ON(fd.id = f.discipline_id)
LEFT OUTER JOIN revisions r ON(r.id = f.file_revision)
LEFT OUTER JOIN record_types rt ON(rt.id = f.record_type_id)
LEFT OUTER JOIN submittal_items_files sif ON(sif.file_id = f.id)
LEFT OUTER JOIN submittal_items si ON(si.id = sif.submittal_item_id)
LEFT OUTER JOIN submittals s ON(s.id = si.belongs_to_submittal_id)
LEFT OUTER JOIN files_actions_groups ffg ON(ffg.file_id = f.id)
LEFT OUTER JOIN groups g ON(g.id = ffg.group_id)
ORDER BY f.created DESC, f.id DESC

Answer 1

This might be an obvious answer, but have you indexed your database? If you're new to indexing, here's a pretty good rule: just put a unique index on all the columns named "id", such as folders.id or projects.id, then put a standard index on all the columns that reference a foreign id, such as folder.belongs_to_folder_id or folder.record_type_id
Another thing I would change is to try and select only the columns you will actually use rather than your huge list of f.*, ts.*, fo.*, ft.*, p.*, u.*, c.*, co.*, ct.*, fs.*, etc...
You also have TONS of joins, which are very expensive in terms of processing time. Do you really need all those joined tables?