Benchmark Factory for Database 8.3 - User Guide

TPC-B Benchmark

• Overview
• Certification of Transaction Processing Council (TPC) Testing Results
• Best Practices
• Scaling Factor

To learn how to create a TPC-B benchmark test in Benchmark Factory, see Create Industry Standard Benchmark Test.

Overview

The Transaction Processing Council is an organization that establishes transaction processing and database benchmark standards. Find a complete overview and detailed explanation of the TPC-B  Benchmark, at: http://www.tpc.org/tpcb/default.asp

Certification of Transaction Processing Council (TPC) Testing Results

Transaction Processing Council  testing results cannot be published as certified unless the testing procedure is audited and approved by the TPC organization. If not certified, the testing results can be published as a "TPC-B like" test.

Best Practices

The following provides best practices for the TPC-B Benchmark.

Load-testing against production servers

Do not load-test against a production server if possible.  Load-testing and benchmarking on a production server significantly degrades performance.  In some cases, load-testing can cause a server to fail.  However, if testing against a production server, take the following precautions:

• Perform the testing when no other users are on the system and no automated processes are running.  Users and automated processes can adversely affect testing results
• Have a recovery plan and backup all data prior to testing
• Determine how long it will take to restore a production server if it went down during load-testing
• Perform manual testing.  Manual testing ensures that no unexpected outside activity takes place during the testing process

Reinitialize the Database

To reinitialize a testing database, run a job containing a Benchmark Object node.  

There are two ways to create a Benchmark Object node:

• Run the Benchmark Object Wizard to add the Delete Benchmark objects for TPC-B node to a job as needed.  
• Run the Load Scenario Wizard to create a new script containing the Create Objects for the Benchmark, Associated Load Scenarios for the Benchmark, and Delete Objects for the Benchmark. Running the delete objects for 'TPC-B' job will clean the environment.

History Tables

The TPC-B benchmark is made up of only one transaction that updates three tables and inserts a record into a history table.  Inserting one record into one history table limits testing performance. The Benchmark Factory properties page allows the user to set the number of history tables to create during a test.  The best ratio of history tables to virtual users is based on database configuration and hardware.  The number of history tables to use is determined by the tester.

Scaling Factor

The TPC-B benchmark scales by a factor of one.