System Sizing Guidelines for Integrity Virtual Machines Deployment -- Hardware Consolidation with Integrity Virtual Machines

10
Depending on the types of workloads and the Integrity server you are using for consolidation, you
can set your target higher (or lower).
Use the high end of your target utilization range for CPU consolidation and the low end for I/O
(both network and mass storage) consolidation. Consider higher thresholds (beyond the upper
range) for workloads that are computationally intensive and have very little interaction with the
operating system or I/O systems. Integrity VM uses the HP-UX fair-share scheduler (FSS) technology
to allocate CPU resources. This technology has been in production for several years as a core
component of HP‟s Process Resource Manager (PRM) and Workload Manager (WLM) products. Its
efficiency has steadily improved over the years and adds very little overhead to CPU-intensive
workloads.
Virtualization and sharing of I/O hardware by Integrity VM (or any other virtualization solution)
will have some impact on performance. Rerouting transactions from the virtual device to the
physical device may require more time and more resources. Because of this, the utilization
threshold for I/O utilization (both mass storage and network) must be more conservative and a
utilization threshold on the low end of your target range is recommended. If extensive resource
sharing is acceptable (e.g., in disaster recovery), then a higher threshold may be considered. Even
so, that threshold should not exceed the high end of your target range.
Aggressively Increasing Hardware Utilization
Integrity servers with higher physical CPU counts may be candidates for higher CPU utilization
thresholds. Marginal increases are appropriate for servers with four to eight physical CPUs; they
can be higher for servers with more than eight physical CPUs. For example, using 50% as a
baseline, consider thresholds of 60% for servers with four to eight CPUs and a threshold
approaching 70% for servers with more than eight physical CPUs.
If you choose to use aggressive hardware utilization ranges, consider using HP Global Workload
Manager (gWLM) to create resource allocation policies – including performance targets relative to
priorities – to workloads executing in the virtual machines. The gWLM can automatically reallocate
CPU resources to the virtual machines in order to meet service-level objectives defined in the
workloads‟ policies. Thus, even during periods of extremely high utilization, gWLM guarantees
high priority workloads access to the necessary resources to perform at the desired performance
level.
The gWLM can also help address periods of higher CPU resource utilization by managing HP‟s
Instant Capacity products. The gWLM product allows you to assign a policy to the VM Host server
to activate Instant Capacity CPU resources whenever hardware utilization exceeds the specified
level.