How to optimize a MySQL JOIN Query

Question

I have this MySQL query that I want to optimize:

SELECT r.WarehouseLocation,sum(sir.qty) 
FROM repairableissue as r 
INNER JOIN SIR ON r.sirno=sir.sirno 
    AND r.region=sir.region 
    AND r.ItemName=sir.Itemdesc 
    AND r.SerialNo=sir.Serialno
WHERE r.status='Pending' 
GROUP BY r.warehouseLocation

How do I optimize this query? I read about optimization and found out that indexes might help but still could not achieve the desired performance.

Which index should be used and which should be removed?

Below is the explain of query:

Repairableissue

CREATE TABLE `repairableissue` (
 `Vendor` varchar(40) NOT NULL,
 `ItemName` varchar(200) NOT NULL,
 `SerialNo` varchar(50) NOT NULL,
 `person` varchar(200) NOT NULL,
 `siteid` varchar(10) NOT NULL,
 `invuser` varchar(50) NOT NULL,
 `region` varchar(50) NOT NULL,
 `id` int(11) NOT NULL AUTO_INCREMENT,
 `Dated` date NOT NULL,
 `Sirno` varchar(50) NOT NULL,
 `status` varchar(30) NOT NULL DEFAULT 'Pending',
 `trackthrough` varchar(30) NOT NULL,
 `reason` varchar(100) NOT NULL,
 `ckh` timestamp NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
 `WarehouseType` varchar(20) NOT NULL,
 `WarehouseLocation` varchar(20) NOT NULL,
 PRIMARY KEY (`id`),
 UNIQUE KEY `id` (`id`),
 KEY `I1` (`status`),
 KEY `ind2` (`ItemName`),
 KEY `ind3` (`region`),
 KEY `ind5` (`SerialNo`),
 KEY `ind4` (`Sirno`)
) ENGINE=MyISAM AUTO_INCREMENT=63029 DEFAULT CHARSET=latin1

sir

CREATE TABLE `sir` (
 `SirNo` varchar(50) NOT NULL,
 `SiteId` varchar(80) NOT NULL,
 `Vendor` varchar(70) NOT NULL,
 `Type` varchar(15) NOT NULL,
 `ItemDesc` varchar(200) NOT NULL,
 `ItemCode` varchar(25) NOT NULL,
 `SerialNo` varchar(50) NOT NULL,
 `Unit` varchar(15) NOT NULL,
 `AssetCode` varchar(50) NOT NULL,
 `Qty` decimal(11,0) NOT NULL,
 `Region` varchar(15) NOT NULL,
 `Status` varchar(20) NOT NULL DEFAULT 'Installed',
 `FaultInfo` varchar(100) NOT NULL DEFAULT 'date()',
 `chk` timestamp NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
 `Phase` varchar(15) NOT NULL,
 `Category` varchar(200) NOT NULL,
 `Issue_Vendor` varchar(30) NOT NULL,
 `AssetName` varchar(150) NOT NULL,
 `Ownership` varchar(20) NOT NULL,
 `Dated` date NOT NULL,
 `PersonName` varchar(150) NOT NULL,
 `Remarks` varchar(300) NOT NULL,
 `po` varchar(100) NOT NULL,
 `invuser` varchar(50) NOT NULL,
 `id` int(11) NOT NULL AUTO_INCREMENT,
 `grnno` varchar(30) NOT NULL,
 `WarehouseType` varchar(20) NOT NULL,
 `WarehouseLocation` varchar(20) NOT NULL,
 `mainpartserial` varchar(200) NOT NULL,
 PRIMARY KEY (`Vendor`,`Type`,`ItemCode`,`ItemDesc`,`SerialNo`,`Ownership`,`SirNo`,`Region`,`WarehouseType`,`WarehouseLocation`,`po`,`Qty`,`id`),
 KEY `id` (`id`),
 KEY `ind4` (`ItemDesc`),
 KEY `ind6` (`SerialNo`),
 KEY `ind7` (`SerialNo`)
) ENGINE=MyISAM AUTO_INCREMENT=228007 DEFAULT CHARSET=latin1

Answer 1

One multi-column index on r.status + r.warehouseLocation, in that order.

One multi-column index on sir.sirno + sir.region + sir.Itemdesc + sir.Serialno, in order of most cardinality to least cardinality, with sir.qty tacked on the end.

This assumes the fields are small enough to fit (combined) into an index.

Still, join seeks are unavoidable. The number of records that match r.status='Pending' is going to dictate the speed of this query.