We will introduce below the database structure in which the development made and tested.
The development and the testing was made in the following databases and its versions:
Postgresql : 8
Db2 : 9
Mssql : 10
Oracle : 11
The development database has the following structure:
In the tables above we tried to make a table, data and key structure from the real world.
On the figure we can see the prodsch2.all_typed_columns table, which tells us that it is the developer database of the Postgresql implementation.
This is the table that contains all of the elementary datatypes supported by the RDBMS to test the data selecting into the test tables.
This table is also defined in Oracle, Mssql and Db2 databases with different columns according to all of the supported datatypes.
All of the other tables have the same structure as the Postgresql tables (because of differently supported datatypes, there may be a few differences).
All of the keys and the data are also the same in all of the implementations.
We inserted a few, 20-20 records into these tables.
The main goal was not the speed check at first but the correctness and the ability to check the data by eyes.
After the running of our automated tests, the result was also checked by hands.
Whether the data transfer of the real data, the larger and binary data was made correctly.
And then.
We have generated 18 million records into this developer database.
As a result, we now have a 4-5GB database.
Creating a 31.83% database from the original (with export and import too but depersonalization not included), takes:
Oracle: 111 minutes
Mssql: 39 minutes
Db2: 76 minutes
Postgresql: 46 minutes
The environment:
OS: Windows 10 Home x64
RAM: 8 GB DDR3 1333 MHz
CPU: 2x2.16 GHz, 1MB L2
HDD: SSD
Our own experience about the usage of the Production Database Reducer in real environments:
From a lot of 100 millions records database > a few thousands of records test database: within about 3 hours.
From about 32 millions records database > about 1.8 millions records test database: within about 2 hours.
The results above are from database servers of a multinational company and from Oracle environment.
As the results of the aboves, got 2 valid testing databases.