Key Partition Performance Measurement

[Imran Pathan]

Measured performance on key partitioned tables and normal tables separately. But we couldn't find any performance improvement with partitioning. Queries are pruned.

Using MySQL 5.1.47 on RHEL 4.

Table details:
UserUsage - Will have entries for user mobile number and data usage for each date. Mobile number and Date as PRI KEY.
UserProfile - Queries prev table and stores summary for each mobile number. Mobile number PRI KEY.

CREATE TABLE `UserUsage` (
`Msisdn` decimal(20,0) NOT NULL,
`Date` date NOT NULL,
.
.
PRIMARY KEY USING BTREE (`Msisdn`,`Date`)
) ENGINE=InnoDB DEFAULT CHARSET=latin1
PARTITION BY KEY(Msisdn)
PARTITIONS 50;


CREATE TABLE `UserProfile` (
`Msisdn` decimal(20,0) NOT NULL,
.
.
PRIMARY KEY (`Msisdn`)
) ENGINE=InnoDB DEFAULT CHARSET=latin1
PARTITION BY KEY(Msisdn)
PARTITIONS 50;

Second table is updated by query select and order by date in first table in a perl program, query is

select * from UserUsage where Msisdn=<number> order by Date desc limit 7
[Process data in perl]
update UserProfile values(....) where Msisdn=<number>

explain partition for select shows row being scanned in a particular partition only.