Problem definition
Right-sizing attempts to properly allocate resources to VMs -- balancing performance requirements with utilization goals. (Allocate too little, and the vM will be starved for resources; allocate too much and utilization drops.)
Many right-sizing solutions today look at a very small set of metrics when right-sizing, and far short of an optimal setting. For example, relying on a single resource metric for VM right-sizing decisions can be misleading. This may have been OK in the "old days" of silo'd servers -- where an application was statically assigned to a physical machine, and the decision of how much resources the application needs was "easy." If it uses only 10% of the CPU, it probably can run on a "slower" machine. If it uses 99% of the CPU, it probably needs a "faster" machine.
This does NOT fly in a virtualized environment, where the VMs are sharing the underlying physical infrastructure, and the performance of a VM may be affected by other entities. For example, maybe the CPU consumption of the VM is/was low because it is/was swapping memory (competing for memory with other VMs), or waiting for I/O because of I/O interference from other VMs.
Reducing the CPU allocations for the VM in cases such as these would be wrong! Or, increasing the memory size of a VM simply because it is consuming 99% of its allocated memory, without first analyzing the overall memory consumption in the underlying physical machine and/or the swapping and ballooning situation, may lead to unwanted results. Resizing/rightsizing a VM is just one possible action to consider to properly provision the environment, BUT it must be looked at holistically, in the context of managing the entire data center. Maybe the optimal action is to move a VM, provision a new physical machine, increasing the physical machine memory, etc.
It is less about rightsizing VMs, and more about rightsizing the data center to deliver the required service. And to do this the right way, one needs to look at a broad range of metrics across the virtual machines, the physical machines, the storage and the applications, understand the VMs/applications priorities, and based on that derive the proper action to perform. And resizing a VM might be (just) one of those actions.
(This, by the way, is exactly why VMTurbo looks at a broad set of real-time metrics and historical trends, and considers a large menu of actions to optimize a virtualized environment.)
Using VMTurbo to right-size your environment
As the ONLY real-time management solution for the virtualized data center, VMTurbo: Pinpoints the problem; Identifies its impact; Recommends the corrective actions; and Can execute them to ensure your virtualized environment is running optimally.
VMTurbo's rightsizing solution identifies the optimal configuration for VMs in real time, and provides reports to quickly share sizing recommendations with stakeholders.
Let's take a look how the four modules work in concert to address this challenge.
Monitor
The Monitor helps you understand whether there are virtual machines, servers, and datastores with high utilization indices -- identifying candidates for right-sizing. And, drilling down into the Problem Log provides further insight to round out the picture, helping you make a better assessment.

Reporter
The Reporter provides historical insights into how long the resources have been over- or under-utlized. By looking at the Top & Bottom Capacity reports, for example, you can determine which VMs are over- or under-sized with respect to specific resources (e.g., Memory, CPU, etc.)
Planner
The wizard-based Planner provides hardware templates to explore a variety of upgrade scenarios identified for the over- and under-utilized server and storage resources. With the Planner, you can look at the overall balance, siff you need more or less hardware of a given size, or would benefit from up- or down-grading.
Optimizer
The
Optimizer looks holistically at your entire infrastructure, and offers a broad set of actions to fix problems and right-size your environment. The list of actions and recommendations include the following:
- Virtual Machine:
- Move between hosts
- Move between datastores
- Start
- Stop
- Suspend
- Change # of CPU
- Change configured vMEM
- Change thin provisioned storage
- Provision required network
- Provision required data store
|
- Physical Machine:
- Start
- Stop
- Suspend
- Provision
- Decommission
- Datastore:
- Provision
- Decommission
- Reconfigure HBA: throttle I/O bandwidth
- Etc.
|
Here's a deeper drill-down on VMTurbo's Optimizer: