Software applications that recognize partition mobility
Software applications might be designed to recognize and adapt to changes in the system hardware after being moved from one system to another.
Most software applications running in AIX®, IBM® i, and Linux® logical partitions will not require any changes to work correctly during active partition mobility. Some applications might have dependencies on characteristics that change between the source and destination servers and other applications might need to adjust to support the migration.
PowerHA® (or High Availability Cluster Multi-Processing) clustering software is aware of partition mobility. You can migrate a mobile partition that is running the PowerHA clustering software to another server without restarting the PowerHA clustering software.
- Software applications that use processor and memory affinity characteristics to tune their behavior because affinity characteristics might change as a result of migration. The application's functions remain the same, but performance variations may be observed.
- Applications that use processor binding will maintain their binding to the same logical processors across migrations, but in reality the physical processors will change. Binding is usually done to maintain hot caches, but the physical processor move operation will require a cache hierarchy on the destination system. This usually occurs very quickly and should not be visible to the users.
- Applications that are tuned for given cache architectures, such as hierarchy, size, line-size, and associativity. These applications are usually limited to high-performance computing applications, but the just-in-time (JIT) compiler of the Java Virtual Machine is also optimized for the cache-line size of the processor on which it was opened.
- Applications that are tuned for given cache architectures, such as hierarchy, size, line-size, and associativity.
- Performance analysis, capacity planning, and accounting tools and their agents are usually migration-aware because the processor performance counters might change between the source and destination servers, as might the processor type and frequency. Additionally, tools that calculate an aggregate system load based on the sum of the loads in all hosted logical partitions must be aware that a logical partition has left the system or that a new logical partition arrived.
- Workload managers.