How Capacity Planning Changes in the Cloud

Written by Joe Kozlowicz on Tuesday, September 20th 2016 — Categories: Cloud Hosting, Cloud Storage

cloud can be added quickly with proper capacity planningTwo of the biggest buzzwords thrown around when talking about cloud are “scalability” and “on-demand.” Those concepts also have implications for your capacity planning as an IT department. You may think that cloud machines nullify the need for capacity planning – after all, if you can just adjust resources on the fly and add or remove processing power and storage as needed, why bother projecting demand?

While it’s true that you can scale as needed, you need to maximize your IT budget and use those dollars efficiently at all times, while avoiding cloud sprawl. Pay-as-you-go only works when you keep a careful eye on your resources, or it can add up quickly when you have unused resources. Capacity planning still has a role to play in your cloud plans.

In a pre-cloud world, even if you planned to virtualize, you would plan capacity by measuring your resource utilization, projecting it out five years or so, and buying enough servers, storage, and network equipment to meet that demand. As part of that projection, you would count on the user groups you know and the applications they use every day. With public cloud resources, those user groups aren’t always just your users. Everything is shared as part of the multitenant model.

Rather than plan for the eventual, and theoretical, peak usage of your applications and user data, cloud lets you adjust IT resources to meet a constantly shifting baseline. The bottleneck therefore comes more from provisioning and configuration rather than the physical hardware in your on premise data center.


1) Systemically Track and/or Limit Resource Consumption

vCenter and vSphere have tools you can use to track and model resource consumption over time, so you know when you might need to provision additional VMs or add vCPUs/RAM/storage. You can also configure alerts to notify administrators when resources are getting stretched thin.

Shares, limits, and reservations are all ways to divide a resource pool within vSphere. You can limit resource consumption by using shares and limits within vSphere, setting RAM limits for 4GB on a VM, for example, even if your virtual data center has an 8GB capacity. However, this is tricky because if you change your provisioning you have to reset the limits, or else you’ll end up paying for unused capacity or may run into performance problems as your VM tries to reach for the additional resources unsuccessfully.

A share is the priority of one VM or pool over another – so if one VM has higher performance requirements, setting it as a higher share means it will pull more resources before the other VMs in the pool do, as it has higher priority.

Finally, reservations are a way to set a minimum resource usage. This is a more flexible way to use all of your allocated resources at the same time. If you have three virtual machines and 4 GB of RAM to go around, with VM 1 and VM 2 each consuming 1.5 GB, then VM3 can use 2.5 GB.

By managing resource pools carefully, you can plan for capacity by pushing more resources towards high priority pools with business critical or high performance applications, while relegating testing and development or less needy VMs to a low performance pool.

White Paper

Need a Managed Security Solution?

Talk to one of our experts today.

2) Consider Implementing Chargeback

If you have multiple IT departments competing for official cloud resources provisioned by IT, you may want to consider implementing chargeback as a way to encourage efficient consumption. In this type of system, each department using cloud resources is charged according to how much they are using. This helps end users or their IT administrators understand what activities lead to higher resource consumption while allowing the primary IT department to see which departments and factors are driving costs.

VMware offers some tools to help implement chargeback and improve IT transparency, like the vRealize Business Enterprise suite.

Chargeback isn’t necessary for cloud capacity planning, however, and it can be a political battle for organizations who haven’t used a similar model in the past. The other difficult part of chargeback is setting a price for different performance tiers or preconfigured virtual machines, depending on the pricing model of your cloud provider.


3) Document and Train Workflow

Workflow management becomes essential to cloud capacity planning and implementation, as delaying the scale of your virtual machines directly affects business efficiency. If your performance is suffering but your workflow is inefficient, that struggling VM is going to keep slowing down end users until you can remedy it.

Training your administrators on efficient provisioning and having a good relationship with your cloud provider can smooth this out (assuming they don’t have a self-service platform that allows real-time provisioning without contract modifications.)

New automation tools including software-defined data center technology, service catalogs, template VMs, and cloning can all be used to quickly implement new VMs.


One recent study found that only half of cloud computing capacity was being used. Those are resources that have already been paid for, sitting idle. Then, due to fear of overpaying, many of the same users then underprovision during peak usage. Without real-time monitoring and careful workflow, these cloud capacity mistakes are all too common.

It takes a mental - and operational - shift to move from a capacity planning mentality focused on supply and projected ceilings to one focused on consumption.