select from merge quicker than individual tables

[Gordon Marshall]

I have created identical tables a_action_log and b_action_log and then created a merge table called action_log. All inserts are via the merge table, the method is last, and therfore into b_action_log. a_action_log has 28121026 rows whilst b_action_log currently has only 23424 rows.

A select from a_action_log (using indexes) takes 40 seconds, whilst the same select from action_log (the merged table) only takes 3 seconds. The select was along the lines of:

SELECT action_type, COUNT(*) FROM action_log
WHERE client=%s AND time BETWEEN %s AND %s
GROUP BY action_type;

with indexes on (client, time) and (client, action_type, time) present.

Would anyone be able to explain why?

Thank you,