Relocation concepts

A z/VM® SSI cluster provides a virtual server mobility function called live guest relocation (LGR).

A running virtual server (guest virtual machine) can be relocated from one member to another. Relocating virtual servers can be useful for load balancing and for moving workload off a member system that requires maintenance. After maintenance is applied to a member, guests can be relocated back to that member, thereby allowing maintaining z/VM and to keep Linux® virtual servers highly available.

The LGR function of z/VM is designed such that the performance impact on the relocated virtual system is minimized. Ideally, the guest operating system running within the virtual system and its applications to their clients should seem to run continuously; this should be the case regardless of the load situation of z/VM, of the guest operating system and of its applications. In other words, the impact resulting from the live guest relocation operation that are noticeable by external clients should be minimized. Particularly, no client visible error situations should result from a live guest relocation.

The "quiesce time" is a primary metric for the live guest relocation process. It is defined as the small amount of time during which – while the live guest relocation operation continues – the virtual server is effectively stopped.

Other relocation related metrics observable from z/VM are the consumption of resources needed for the relocation process, such as the consumption of CPU and memory, and the I/O rates and data amounts processed by the ISFC logical links, and their FICON® CTC devices and FICON channels. Of course the relocation process is impacted if any of these resources is limited.

A metric observable from the perspective of a Linux guest operating system is the steal time. The steal time is the percentage of time that the guest operating system is involuntary wait state while z/VM is doing something else, such as for example performing a live guest relocation.