Concurrently update the scan-initialization data of a processor core

• Authors: C. R. Conklin;W. Niklaus;R. Schaufler;S. Swaney
• Venue: IBM Journal of Research and Development
• Year: 2012

The IBM zEnterprise® 196 (z196) server has the ability to update the scan-initialization data of a processor core concurrent to the operation of the server. In previous System z® servers, updates of the scan-initialization data of a processor core could only be performed disruptively, requiring a customer to again perform an initial microcode load on his server to pick up changes to the processor core scan-initialization values. With the z196 server, a spare processor core is clock stopped, the new scan-initialization data is applied, the clocks on the spare processor core are then restarted, and the spare processor is brought back into the server configuration concurrent to the functional operation of the server. To achieve the concurrent application of the scan-initialization data to a nonspare processor core, the physical backing of this nonspare processor is first swapped with a spare processor, the updates are made, and then the physical backing is swapped back in order to maintain the original physical configuration. This process is invoked on each physical processor core in the system until all processor cores have been updated. Handling of updates of the scan-initialization data is smoothly embedded into the existing "fix application" framework so that dependence with other code updates is transparently handled for the customer. With this new capability, both "fix application" and the possibility of enabling new hardware functionality on a processor core can now be done concurrently with the server operation, which is another step toward the goal of achieving 100% availability in a System z server.