HP Matrix Operating Environment Software - Manual (Page 107)

Several predefined queries provide ways to look for underused systems.
Table 9 Predefined analysis queries: Finding underused systems
Description Query name
Physical servers having little to no resource utilization over the
selected interval.
Consider removing these systems or converting them to VMs.
Low usage hardware
A VM's peak memory utilization is significantly less than what is
allocated. Allocating too much memory to a VM that does not need
it can prevent other VMs from being run.
Consider reducing the memory by half.
VMs with over-provisioned memory allocations
VMs with a peak CPU utilization significantly less than what is
provisioned.
Consider removing half of the vCPUs from these systems to reduce
overhead in the hypervisor and possibly reduce the licensing costs
of the software running in the VM.
VMs with over-provisioned vCPUs
Physical peak memory utilization significantly less than what is
provisioned. It is probably safe to remove half of the memory, or
move this workload to a smaller server. Systems with over-provisioned memory allocations
VMs that have had little to no resource utilization over the interval.
These VMs could be deleted, making room for more useful VMs. Low usage VMs
This guide also includes a custom query example that finds underused systems. See “Example:
Finding large underused VMs that could be replaced with medium or small VMs” (page 122) .
Finding overused systems
Capacity Analysis can identify servers that are overused over a longer time horizon. When CPU
utilization is above 70 or 80%, queuing delays for interactive workloads can exceed the time the
CPU spends doing useful work. Response times can double or triple long before CPU use reaches
100%. Searching for, and fixing, these more subtle cases of system overuse now can prevent crisis
situations later.
The predefined queries below provide ways to look for overused systems.
Table 10 Predefined analysis queries: Finding overused systems
Description Query name
Shows systems with a Five-Star rating equal to 0. These systems
may not be meeting their Service Level Agreements. It could be that
the systems need more resources.
Edit the utilization limits in the Configure menu of the Workload
tab.
See “Headroom star rating” (page 26) for more information about
the star ratings.
Systems exceeding their utilization limits
Systems with memory paging rates greater than 500 pages/second.
Consider adding memory to these systems, or moving the workload
to a larger system.
Systems with a high paging rate
Systems with a 90th percentile CPU run queue greater than 5. A
CPU run queue higher than 2 generally indicates a performance
issue.
Consider adding more processors to these systems.
High CPU queuing
Predefined analysis queries 107