Another consideration for server farm operations is failover. Because z/VM controls the hardware, it can be seen as a single point of failure. Therefore, you might want a failover system in case z/VM suffers an outage. Depending on company policy, there are three possibilities for a backup system: a duplicate image, a duplicate z/VM on another LPAR, or duplicate machines. To maintain service during a system failure, company policy might state that a duplicate server must be available to back up the primary server. A physical hot-standby configuration duplicates acquisition and environmental costs for a server that you hope will never have to be online. With z/VM, the costs associated with a backup Linux virtual machine waiting to become active are minimal. If a hot-standby server in the form of a duplicate image is activated, it simply uses the system capacity relinquished by the primary server that just failed. zSeries 800 or 900 machines can be partitioned into up to 15 logical partitions using the PR/SM feature. One partition can contain a z/VM that acts as a hot standby for the z/VM system in another logical partition. This may require some custom programming (for example, developing a "heartbeat" monitor to verify the operational status of the systems, and programmable operator scripts to initiate procedures on the standby system). Linux-based failover solutions can also incorporate a secondary z/VM system as part of the approach. See Chapter 11, "Achieving Higher Availability," for more information. |