Capacity Planning
VM performance and capacity are linked. Without sufficient capacity, applications suffer performance problems. Without accurate capacity planning, a growing environment suffers from firefighting, rushed procurement cycles, and poor ROI on virtual infrastructure.
vOPS Overview Video
Avoiding VM Performance Problems Requires Capacity Planning
Ensuring that sufficient capacity will be available in a virtual environment to support further VM growth and to maintain high performance is critical for virtualized data center operations. Because VM demand for resources is unpredictable, virtualization administrators must be able to access realistic assessments of available capacity in an environment, and model growth scenarios to determine procurement needs. Also, VMs left to their own operations will grow, requiring more resources, and can cause capacity bottlenecks that impact performance if insufficient capacity is available.
Virtualization administrators can certainly wait for these problems to manifest themselves and then react. This causes
- Unbudgeted hardware procurement
- Excessive firefighting
- Dissatisfaction from end users
- Loss of credibility that an environment is truly enterprise ready
For many organizations, a long procurement cycle may make short term fixes for VM capacity problems dependent on additional hardware installation impossible. Clearly, taking steps to ensure VM capacity for growth and avoid VM performance problems before they manifest themselves is critical to improving customer satisfaction, reducing firefighting hours, and predictably acquiring new hardware on a planned, not panicked basis.
Capacity Planning Accuracy Prevents Environment "Growing Pains"
How much Hyper-V or VMware capacity is available today to slot new VMs? How much hardware will be needed in the next purchase cycle for the environment? Virtualization administrators require this information as a baseline to plan for future expansion, add resources where needed, increase capacity and perform best placement for new applications. These actions must be undertaken while maintaining the optimal VM density for a host. The result is lower capital costs, operating costs, and improved service levels as new VMs can be quickly slotted into available space.
Determining available Hyper-V or VMware capacity is not a simple calculation however. Available resources don’t translate to number of VMs that can be placed. To correctly translate available CPU, memory, storage, throughput to available VM slots, some model of VM size needs to be applied. “Average” VM models are a good starting point but rarely work beyond simple environments since most organizations have multiple VM sizes. To successfully determine available capacity, any capacity management system should be able to use custom VM models specified at the host or resource pool level or based off of existing VM's resource usage. Whichever model is used, all available Hyper-V or VMware capacity calculations must take into consideration maintaining high availability requirements imposed at the cluster level.
Effective Capacity Planning and Issue Forecasting Requires Analytics
Predicting capacity needs for environment growth and avoiding VM performance problems resulting from resource shortages requires historical analysis of 20 VM metrics at the VM, host, cluster, and resource pool level over a 30-day period. A future consumption prediction must then be made for each VM, then rolled up across all VMs in a host, cluster, resource pool and data center.
Using a product like Microsoft Excel can simplify the data collection process, but linear analysis over-simplifies the calculations required to predict future problems. Predicting future issues and hardware needs is a complex mathematical problem that can only be solved accurately with advanced analytics. Properly solving the problem involves predicting not only when the performance bottleneck will occur but also determining the constraining resource or resources. Identifying the constraining resource is critical to either rebalancing the environment or adding hardware. Without identification of the constraining resource, virtualization administrators are left to guess as to whether they have a problem that can be solved by rebalancing or whether they need to procure more hardware or hardware components. Additionally, because VM demand is unpredictable, modeling multiple growth scenarios and removing capacity that will be used for planned deployments is essential to accurately predict how much capacity will be needed in the next budget cycle.
Finally, to properly avoid capacity bottlenecks, some measure of capacity alerts are required to flag potential problems on the basis of either rising average utilization levels or predictions of out of band utilization. Properly implementing a predictive bottleneck use case can reduce CAPEX on new equipment, reduce firefighting times, and improve end user satisfaction.
VKernel Capacity Manager Forecasts and Manages Virtual Environment Capacity With Analytics
VKernel vOPS Capacity Manager, part of the VKernel vOperations Suite, shows virtualization administrators when performance issues will occur due to capacity constraints, and how many more VMs can fit in the existing Hyper-V or VMware environment under multiple growth scenarios. Using the patent pending Capacity Analytics EngineTM, Capacity Manager offers mathematically precise results to ensure accurate capacity planning. Installed as a virtual appliance, Capacity Manager provides value within 20 minutes from download by:
- Visually showing the capacity status of all hosts and VMs in an environment across multiple data centers and hypervisors
- Predicting future performance issues and the constraining resource that will cause the problem
- Forecasting future CPU, memory and storage requirements for any point in the future
- Determining available VM slots based on average or maximum VM resource usage, existing VM's resource needs, or for a customized VM’s CPU, memory, storage, and disk throughput measures. Individual datastores can be assessed for storage and disk throughput capacity
- Increasing VM density by calculating the maximum amount of VMs that will safely fit on a host
- Modeling the number of VM slots that will be available in an environment’s future state in multiple scenarios
- Reserving capacity needed by VMs that will be deployed in the future to remove those resources from further modeling
- Providing capacity alerts based on storage commitment
- Identifying top resource-consuming VMs
- Allowing for capacity planning data integration into external systems such as self-service portals and orchestration engines
Planning for Hyper-V and VMware capacity with Capacity Manager results in performance issue avoidance, higher VM to host consolidation ratios and accurate hardware procurement forecasts. Capacity Manager is simple to download, easy to deploy and provides immediate measurable value.
