IBM Systems Magazine, Mainframe Edition - November/December 2010 - (Page 40)

Tech Corner Programming, systems operations and more A Soft Touch z/OS Predictive Failure Analysis makes it easy to spot and fix soft system failures By Karla Arndt, James Caffrey and Aspen Payton W hen OS/360* was released in 1964, data processing focused on automation of back-office processing like payroll. Today, solutions using z/OS* often represent your organization’s public face. To meet the increasing demand for resiliency caused by this change, the z/OS team delivered Parallel Sysplex*, Geographically Dispersed Parallel Sysplex* (GDPS*), HyperSwap*, and in z/OS V1R10: Predictive Failure Analysis (PFA). PFA extends z/OS resiliency by detecting “sick, but not dead,” or soft failures. Failures on z/OS can be divided into three categories: hÁMasked failures: the software detects and corrects the failure with no impact to business hÁ Hard failure: the software detects the failure, but automation and operations restart the failing component with minimal impact to business h Soft failure: users detect soft failures, thus business is significantly impacted; it’s difficult to determine recovery actions With soft failures, a rare combination of legal, but abnormal, behavior causes the system to externally appear to stop working. For example, a subsystem experienced tens of fail- ures each minute, which continued for days without a problem until the workload on the system approached 100-percent utilization. At that time, work on the system appeared to stop. This event could’ve been easily mitigated if the operations team recycled the subsystem before the system approached its maximum utilization. Infrequent Problems, Big Business Impact Data analysis from some of our largest customers indicates soft failures only account for about 20 percent of system failures, but can cause 80 percent of the business impact. The objective of PFA is to convert these soft failures into correctable incidents to help avoid system outages before they impact business operations and customer transactions. 40 NOVEMBER/DECEMBER 2010

Table of Contents for the Digital Edition of IBM Systems Magazine, Mainframe Edition - November/December 2010

IBM Systems Magazine, Mainframe Edition - November/December 2010
On the Web
Editor's Desk: Tradition Takes Planning
Dashboard: Walk While You Work
Data Display: All About Spam
Think Smarter: IBM Offers Smarter Systems for Performance and Scalability
Trends: Rosamilia Oversees Both System z and Power Systems Lines as New GM
Break Through Economics: Dr. Howard Rubin Discusses Mainframe Efficiencies and the zEnterprise System
Streamlining Development: IBM Rational on zEnterprise System Utilizes Multiplatform Development Capabilities
Administrator: The zEnterprise System Changes Firewall Requirements
Technical Corner: z/OS Predictive Failure Analysis Make It Easy to Spot and Fix Soft System Failures
Developer: Native XML Support Strengthens DB2 and COBOL Development
Advertisers' Index
Stop Run: Former IBMer Jim Bell Finds Inspiration in Music

IBM Systems Magazine, Mainframe Edition - November/December 2010