Mainframe-Class Recoverability Testing

The corollary to the axiom "all software has bugs" is "you will never find them all." Even if you could, hardware and environmental failures always are lurking about, waiting to crash the software. If you accept the premise that failures are inevitable, then part of your testing should confirm that the software gracefully recovers from failures, protecting customer data and minimizing downtime. In this presentation Scott Loveland helps you face the issue head-on by explaining novel ways to force failures and then test the software's ability to recover. Having spent his career with IBM in test for z/OS and its predecessors, MVS and OS/390, and most recently Linux, Scott reveals the tools and techniques proven for testing recoverability of industrial-strength software in the trenches of the IBM mainframe development lab.

• Methods for injecting errors and monitoring recovery of large, complex systems
• Ways to trigger module-level, system-level, and multi-system-level error paths
• How the recovery testing approach varies by testing phase