In my last post, I discussed how vCenter CapacityIQ enables capacity optimization by tracking powered-off and idle virtual machines. In this post, I will cover how CapacityIQ identifies oversized and undersized virtual machines.
Like powered-off and idle virtual machines, CapacityIQ uses configurable utilization
thresholds to identify oversized or undersized virtual machines. However, unlike powered-off and idle virtual machines, CapacityIQ uses a two-level analysis to
mark a virtual machine as oversized or undersized. Each of these two levels has
a separate threshold, which can be set from the CapacityIQ Global Settings. CapacityIQ currently uses CPU and memory metrics for analyzing both oversized and undersized virtual machines. As stated in the last post, you can also customize these views by selecting specific clusters on the left-side inventory tree or by changing the time interval used for the analysis. Let’s start with the oversized virtual machines.
Oversized virtual machines
CapacityIQ
defines oversized virtual machines as those that consistently use less capacity than their
configured capacity. The motivation behind this analysis is to identify virtual
machines that can be right-sized down to a smaller configured capacity. For
virtual machines identified as oversized, CapacityIQ also calculates a
recommended CPU and memory configuration such that right-sizing the virtual
machine to the recommended configuration will result in the virtual machine being
no longer under-utilized.
For
the given time interval, CapacityIQ first calculates if a physical resource (viz.
CPU, memory) is under-utilized based on the configurable Utilization Less Than threshold. Next,
it calculates the degree to which a virtual machine is oversized. A virtual machine is identified as oversized if its degree oversized
is greater than the % Oversized threshold. The
idea behind this two-level approach is to factor both (1) the amount of time a
virtual machine spends in an underutilized state, as well as (2) the extent to
which the virtual machine is underutilizing its configured capacity. This is
illustrated in the following figure. The horizontal Threshold(U) line defines the first Utlization Less Than threshold. The degree oversized is the ratio of the area shaded blue to the area
of the grey box and defines the % Oversized threshold.
Once a virtual machine has been identified as oversized,
CapacityIQ uses a proprietary technique to calculate a recommended configured
capacity such that, the virtual machine will no longer be underutilized if
right-sized to the recommended configuration. The Oversized Virtual Machines List view shows the final result of this analysis.
Undersized virtual machines
The
motivation behind this analysis is to identify virtual machines that might need
to be right-sized up to larger configured capacity so that the workloads
running within the virtual machine can get sufficient capacity. Unlike oversized virtual machines list, however, CapacityIQ does not provide a
recommended CPU and memory configuration for undersized virtual machines.
To
identify undersized virtual machines, CapacityIQ uses similar two-level
analysis as oversized virtual machines. As with the oversized virtual machines analysis, each
of these two levels has a separate threshold, which can be set from the
CapacityIQ Global Settings. For
the given time interval, CapacityIQ first calculates if a physical resource
(viz. CPU, memory) is over-utilized based on the configurable Utilization More Than threshold. Next,
it calculates the degree to which a virtual machine is undersized. Once again, the degree undersized is the ratio of the area shaded blue to the area
shaded grey box, as illustrated in the following figure. A virtual machine is identified as undersized if its degree
undersized is greater than the %
Undersized threshold. The Undersized Virtual Machines List view shows the result of this analysis.
As with Powered-Off and Idle virtual machines, the thresholds for Oversized and Undersized
virtual machines can be configured from the same CapacityIQ Global
Settings screen, shown below. Note that you can configure CapacityIQ to detect utilization based on just one or both the metrics.
