优化查询性能

Question

My Appliation(using php) performance is very slow due to more data in my mysql table, My table Structure as follows,

`CREATE TABLE `xxxxx` 
(`ID` int(11) NOT NULL AUTO_INCREMENT,
 `IP` varchar(45) NOT NULL,
 `Cname` varchar(45) NOT NULL,
 `Port` int(5) NOT NULL DEFAULT'3306',
 `Variable` varchar(150) DEFAULT NULL,
 `status` varchar(100) DEFAULT NULL,
 `moddate` datetime NOT NULL,
 `Servertime` datetime NOT NULL,
 PRIMARY KEY (`ID`,`moddate`),
 KEY `idx_variable` (`Variable`) ) 
  ENGINE=InnoDB AUTO_INCREMENT=85054928 DEFAULT CHARSET=ucs2
    /*!50100 PARTITION BY LIST ( MONTH (moddate))
    (PARTITION Jan VALUES IN (1) ENGINE = InnoDB,
    PARTITION Feb VALUES IN (2) ENGINE = InnoDB,
    PARTITION Mar VALUES IN (3) ENGINE = InnoDB,
    PARTITION Apr VALUES IN (4) ENGINE = InnoDB,
    PARTITION May VALUES IN (5) ENGINE = InnoDB,
    PARTITION Jun VALUES IN (6) ENGINE = InnoDB,
    PARTITION Jul VALUES IN (7) ENGINE = InnoDB,
    PARTITION Aug VALUES IN (8) ENGINE = InnoDB,
    PARTITION Sep VALUES IN (9) ENGINE = InnoDB,
    PARTITION Oct VALUES IN (10) ENGINE = InnoDB,
    PARTITION Nov VALUES IN (11) ENGINE = InnoDB,
    PARTITION DECM VALUES IN (12) ENGINE = InnoDB) */`

And It has around 4.2 GB data. My select statement for fetch data is,

SELECT 
      `status`,
      `moddate` 
   FROM 
      `xxxxxx` 
   WHERE 
          Cname='xxxx' 
      and ip='0.0.0.0' 
      and port='3306' 
      and Variable='xxxxxx' 
      and status REGEXP '^[0-9]+$|^ON$' 
      and moddate between '2015-01-01 00:47' and '2015-02-01 01:07';

but it takes 4:00 min. Since 15 days i have been suffering by this problem. I know we need to put index on column. But I don't know how to optimize this. Could you please any one suggest me which field needs to indexed. or share any other link to optimize my application. thanx...

Answer 1

I would look to the where clause and try to determine what might be the smaller result set field first and have that in the first position of a composite index

You may have many entries for the client "xxxxx", but a smaller set on the IP address. Not sure what your "Variable" context is, maybe this is smaller. But I would try on

( IP, port, cname, variable, moddate, status )

but that is basically an index on ALMOST everything. So, that said, I might chop it down to see performance based on...

( IP, port, cname, moddate )

After thinking more on it, having CName in first position is probably better...

( CName, IP, port, variable, moddate, status )

as it hit me cname is probably like a website address, and since a server on a single IP can host many website names, the CName would be smaller than the overall IP address.

Answer 2

This is your query:

SELECT status, moddate
FROM xxxxx 
WHERE Cname = 'xxxxxx' and
      ip = '0.0.0.0' and 
      port = 3306 and
      Variable = 'TABLE_LOCKS_IMMEDIATE' and
      status REGEXP '^[0-9]+$|^ON$' and
      moddate between '2015-01-01 00:47' and '2015-02-01 01:07';

Note: port is an integer so I removed the single quotes around 3306 . Sometimes, in some databases, the use of mixed comparisons can affect the use of indexes (although probably not in this case).

Your query is accessing one table, so the place to start with defining an index is the WHERE clause. An index that has all six columns mentioned in the table is called a "covering" index. And a covering index is usually the best for performance.

For such an index, start with the columns that have equality conditions. Then you get one position for inequalities and in . Then the rest of the columns are in the index just so the engine doesn't have to access the data pages. So, the best index is xxxxxx(Cname, ip, port, Variable, moddate, status) . The ordering of the first four columns does not make a different for this query, because all have equality indexes in the from clause.