Currently all ppc64 logical partitions are reported by the inventory scan as "bare metal" systems (i.e. no hypervisor value). This is fine for partitions with static resource allocations (since they should be as good as bare metal for basically any purpose you care to name), but can cause problems for partitions with shared resource allocations. The main concern arises when looking for systems to do performance testing - for other architectures, its possible to filter for "bare metal only" systems, and be assured of being the sole tenant on the host. With ppc64, you may end up running on a logical partition with shared resources, and the benchmarking results will be affected by activity on other partitions. This RFE proposes reporting a suitable hypervisor string (e.g. "LPAR") in cases where a ppc64 system is running in a logical partition with shared (rather than static) resources. It assumes that it will be possible for beaker-system-scan to: 1. Detect when it's running in a logical partition 2. Detect when that partition is running with shared resources http://www.ibmsystemsmag.com/aix/administrator/lpar/An-LPAR-Review/ provides a good conceptual overview of LPAR.
"virt-what" (see http://people.redhat.com/~rjones/virt-what/) may provide some useful hints on detecting additional hypervisor information.
The virt-what man page (linked from the above page) shows more details on what it can detect: http://people.redhat.com/~rjones/virt-what/virt-what.txt
Not directly relevant, but https://unix.stackexchange.com/questions/89714/easy-way-to-determine-virtualization-technology has some interesting answers regarding poking around trying to identify the virtualisation layer.