Capacity Management is well established practice in enterprise data centers. Effective capacity management strives to accomplish two objectives:

• Efficiency: Make efficient use of existing capacity and minimize any waste of computing resources
• Availability: Leave enough capacity headroom to accommodate any spikes in resource demands.

These two capacity management objectives often conflict with each other. Rightsizing resource capacity to achieve high efficiency usually comes at the cost of sacrificing capacity availability. Similarly provisioning additional capacity buffer tends to reduce capacity efficiency. Consequently Capacity Management is a balancing act and tends to become more of an art than science.

Capacity Management in virtual environment is more challenging for couple reasons. First, the physical resources in virtual environment are shared across multiple virtual machines. Second, vSphere has its own advanced features, such as DRS and Memory optimizations, which manage resource allocations continuously and dynamically.  Several VMware customers have tried to develop homegrown solutions based on Excel and/ or scripts. Unfortunately these solutions either tend to ignore these nuances in virtual deployments, and often over-estimate capacity needs. Or they provide point-in-time snapshot of capacity and need to run regularly on an ongoing basis. vCenter CapacityIQ is the purpose built tool from VMware that takes into account these nuances and helps you address capacity management challenges in virtual deployments. As the CapacityIQ product page states, it will help you with Capacity Awareness, Capacity Optimization and Capacity Prediction. In this series of posts, I will elaborate on each of these use cases, starting with popular one - Capacity Optmization. As always, you can also download and test fully-functional evaluation versions of CapacityIQ free for 60 days here.

CapacityIQ includes several views to help you optimize the use of existing capacity in the virtual infrastructure:

• Powered-off virtual machines
• Idle virtual machines
• Oversized virtual machines
• Undersized virtual machines

You can customize these lists in several ways e.g.

• select specific clusters by navigating the inventory tree in the left pane
• change the time interval used for the analysis
• configure the settings used to qualify the powered-off/idle/oversized/undersized thresholds.

Powered-off virtual machines

CapacityIQ identifies virtual machines that have been placed in a powered-off state for a significant portion of time. The motivation behind this analysis is to identify virtual machines that are rarely or never used. Such virtual machines are candidates for decommissioning and represent an opportunity to reclaim unused capacity.

CapacityIQ identifies Powered-Off Virtual Machines based on a threshold setting that’s configurable from CapacityIQ global settings. Using the graph below, a virtual machine is identified as Powered-Off in a given interval, if the total time it is powered-off (% Time Powered-Off) is greater than the configured % Time Powered-Off Threshold. See the CapacityIQ global settings screenshot under 'Idle virtual machines'  to see where you configure the % Time Powered-Off Threshold.

Idle virtual machines

CapacityIQ identifies virtual machines that show an insignificant level of capacity usage on one or more resources most of the time. The motivation behind this analysis is to identify virtual machines that have practically fallen out of use for any number of reasons – e.g. users are no longer running workloads within the virtual machine. Such virtual machines are candidates for decommissioning and represent an opportunity to reclaim unused capacity.

CapacityIQ identifies Idle Virtual Machines based on capacity usage of the following resources:

• CPU
• Disk I/O
• Network I/O

For each of these resources, you can set utilization threshold from the CapacityIQ Global Settings. For a given time interval, CapacityIQ calculates the total time during which capacity usage is below a given Usage Less Than threshold.  If the percentage of this total idle time is more than % Time Idle Threshold, then the VM is identified as idle for that particular resource. You can configure CapacityIQ to identify idle virtual machines based any one or all of these thresholds. The following figure illustrates how CapacityIQ identifies Idle Virtual Machines.

The thresholds for Powered-off and Idle virtual machines can be configured from the same CapacityIQ Global Settings screen, shown below.

I will cover the Oversized and Undersized virtual machines views in the next post. As always, let me know any comments, questions you have.