One of the persistent perceptions of cloud computing is that costs associated with on-demand services are too unpredictable. The initial costs entice organizations to adopt more services; however, poor controls and the resulting server sprawl due to the ease of deployment of new servers and services lead to billing shock of monthly cloud computing expenses. The industry has taken notice and has started to provide options that make expenses more predictable.

The most popular infrastructure as a service (IaaS) is Amazon’s Web Services (AWS). In the traditional AWS model, end users paid for cloud services on a per minute basis. This model lowered the entry cost for developers and startups compared to the costs of building an equivalent on-premises infrastructure. Cloud customers quickly found workloads not requiring elastic usage become expensive. Traditional enterprise workloads are not elastic.

An ERP application server is a great example of traditional workloads that may not be suitable for an on-demand pricing model. An ERP application server doesn’t power down when not in use. Therefore, Amazon’s traditional elastic pricing does not lend itself well to traditional enterprise workloads. Couple the pricing model with server sprawl from ease of deployment, and the result is cloud cost overrun. In an effort to address the predictability of IaaS cost, the industry has come up with additional pricing options.

Reserved pools

A common pricing model is to price computing resources based on a pool of resources. Pricing for VMware’s vCloud Air follows this pool resource concept. In the case of VMware vCloud Air, customers are billed for a fixed amount of resources during a billing cycle. The customer has the option and tools to provision the resources across virtual machines as needed. In the case of vCloud Air, the customer even has the option to overprovision their pooled resources. Performance is guaranteed up to the billed amount of resources. Performance will potentially suffer if the customer tries to use more resources than allotted.

Reserved pools give customers some assurance on billing. With the reduced risk comes reduced flexibility in savings on compute cost. There’s a strong chance of waste if a customer overestimates their cloud computing needs. Regardless of the actual usage, the bill remains static. For organizations looking to be as efficient as possible in their cloud billing and usage, pooled resources can be difficult to track.

Reserved instances

Amazon recently announced the availability of reserved instances. Reserved instances allow customers to pre-pay for a set number of AWS instances at a reduced price. Similar to reserved pools, reserved instances allow for predictable cloud billing. In the earlier ERP use case, an organization can pre-purchase the instance for the entire year. In return, Amazon will offer discounts on the virtual machine up to 75% less than the cost of an equivalent on-demand instance.

Just like reserved pools, some flexibility is lost with reserved instances. Amazon’s terms for reserved instances are 1 to 3 years. During this period, Amazon will allow the change of attributes such as availability zone and networking, but the terms of the instance stay static.

Sustained use policy

Google may offer the simplest option. Google uses a sustained use policy to dictate pricing for its Compute Engine service. The longer an instance is used, the cheaper it gets throughout the month. There is no upfront commitment required. Google’s pricing may appeal to organizations looking for a break on on-demand pricing while still having the flexibility of elastic compute.


The cloud industry has begun to respond to the pricing challenges that hinder some enterprises from embracing public or hybrid cloud. There are now solid options from the major vendors that help control the cost associated with server sprawl in the cloud.

Has on-demand pricing stopped you from moving workloads to the cloud? If so, will these new pricing models make you think twice about cloud? Let us know in the discussion.