Capacity Planning in a Virtual World
Intelligently Plan to Meet Future Demand
An activity as important as capacity planning must provide an accurate projection of the expected resource demand levels for each workload; it absolutely has to be performance-based. Additionally, that same planning activity should tell us exactly how (where) to allocate that workload (VMs to physical hosts) and those recommendations should take into account a wide array of parameters that are often overlooked by capacity-based approaches to planning.
The goal of capacity planning is to grow the virtual footprint comfortably, and in many cases use the existing hardware to do so. Understanding the performance footprint, and its historical variations, is absolutely critical to get capacity planning right.
In the virtualized infrastructure this can be very challenging due to the ease with which new virtual machines can be provisioned and the dynamically changing characteristics of workloads which share common server and storage resources. Capacity planning is further complicated by the desire of application owners to adopt best practices advocated by software vendors, which were established for deployment of their applications in physical environments. This can often result in oversizing of virtual machine resources and waste resources.
VMTurbo Operations Manager looks at capacity planning in a fundamentally different way. It uses an economic abstraction based on supply and demand principles to balance the demand from workloads with available server and storage capacity. It applies these principles in real time to constantly optimize workload performance and ensure server and storage resources are used most efficiently. VMTurbo enables you to take this a step further and run offline “what if” planning scenarios to look into the future and answer key questions for critical projects such as:
- Consolidating data centers: Will I need more hardware? What utilization metrics will I see after the merge? Where will bottlenecks arise as additional resources are sharing the new environment?
- Upgrading server hardware: How many of the new servers will I need? What utilization metrics will I see? What happens as load changes over time?
- Improving server packing: How can I better allocate resources across my existing estate to improve performance and free up room to grow virtualized workloads?
- Adding new VMs: Will I need to add more hardware? What utilization metrics will I see after adding the load? What if I add a multiple types of workloads?
- VM right-sizing: Are the requested allocations – CPU, memory, storage – correct for each of the workloads? Is the VM over or under provisioned? How will right-sizing affect performance across my virtual datacenter?