mysql 时间查询优化

mysql按照时间查询不要使用subdate函数，因为每次subdate的结果值都不同，导致无法query_cache:

这条SQL：

select id,add_datetime,favorite_count  from message_message
where add_datetime >  SUBDATE(now(),INTERVAL 25 HOUR)  
and category IN (1, 9) order by favorite_count desc limit 24;

执行三次：

24 rows in set (1.01 sec)
24 rows in set (0.98 sec)
24 rows in set (0.95 sec)

修改成:

SELECT id,add_datetime,favorite_count  from message_message 
where add_datetime > '2012-08-29 13:44:44'  and category IN (1, 9)
order by favorite_count desc limit 24;

执行三次：
24 rows in set (0.98 sec)
24 rows in set (0.00 sec)
24 rows in set (0.00 sec)



不建索引：
select id,add_datetime,favorite_count  from message_message   where add_datetime >  '2012-08-30 12:13:41'  and category IN (1, 9)  order by favorite_count DESC limit 24 ;

24 rows in set (50.44 sec)
24 rows in set (0.00 sec)
24 rows in set (0.00 sec)

explain select id,add_datetime,favorite_count  from message_message   where add_datetime >  '2012-08-30 12:13:41'  and category IN (1, 9)  order by favorite_count DESC limit 24 \G;
------------------------------------------------------
           id: 1
  select_type: SIMPLE
        table: message_message
         type: index
possible_keys: idx_add_t,idx_add_rep,idx_cat_add_fav
          key: idx_favorite_count
      key_len: 4
          ref: NULL
         rows: 48505
        Extra: Using where
1 row in set (0.00 sec)
explain 信息:
           id: 1
  select_type: SIMPLE
        table: message_message
         type: range
          key: idx_cat_add_fav
      key_len: 10
          ref: NULL
         rows: 34899
        Extra: Using where; Using index; Using filesort

idx_cat_add_fav(`category`,`add_datetime`,`favorite_count`)

select id,add_datetime,favorite_count  from message_message   where
add_datetime >  '2012-08-29 13:43:44'     and category IN (1, 9)  order by favorite_count limit 1,10;

这条sql不知道为什么非常慢

执行计划:
Extra Using where; Using index; Using filesort
key:  idx_cat_add_fav
type: range

索引：
idx_cat_add_fav(`category`,`add_datetime`,`favorite_count`)

select count(*)  from message_message   where add_datetime >  '2012-08-29 13:43:44'     and category IN (1, 9)

count: 17400