While cloud computing touts its lower-than-on-premises costs, pricing a transition to the cloud is not that simple. With traditional approaches to IT infrastructure, a manager had to consider many variables, including server and other hardware, when rolling out new software and services. Cloud providers promise to reduce application environment complexity to enable easier buying decisions. It's too bad they don't simplify pricing options, said Owen Rogers, senior analyst of digital economics for 451 Research and author of 451's Cloud Pricing Codex.
The VM is just the beginning
A common mistake cloud-newcomers make is including the cost of virtual machines in their calculations but not other cloud computing costs. The costs of the supplemental services associated with an application, like bandwidth and storage, can significantly increase the overall expense of running a new application.
It's preferable to take a holistic approach that considers the estimated use of all the components for deploying an application in the cloud. Furthermore, these costs can also vary by region, based on the type of server (expect to pay more for Windows versus Linux), time frame and other variables. There are also bundling options that in addition to a processor include RAM, local storage, bandwidth, object storage, load balancing IP addresses and block storage.
These costs are always changing thanks to competing services like AWS, Azure, Google and Rackspace. Consequently, it helps to assess the options using a tool that tracks these constantly changing numbers, much like the stock market. A good starting point for comparing providers is the CloudVertical Cloud Cost Index.
Take measures against cost spikes
In some cases, it might be a good thing that cloud computing costs are ramping up as a result of a growth of legitimate uses. But, in others, spikes could be caused by cyber-attacks or poorly written applications.
The costs of the supplemental services associated with an application, like bandwidth and storage, can significantly increase the overall expense.
The concern is that without proper management, costs can grow out of control.
Cloud providers often provide some ability to control the scaling of VMs to set limits, and managers are wise to take advantage of these. But it is important to apply similar management to bandwidth and storage as well, said Rogers.
Simulating the total cost
Because of the complex nature of cloud pricing variables, it helps to simulate the application(s) that a company intends to deploy to the cloud in order to get a more accurate assessment of the total cost of cloud ownership. It's also a good idea to have a trusted adviser recommend where cost savings can be made during application development.
In addition, a variety of third-party tools have been developed to help compare the total cost of existing, on-premises deployments to moving these to the cloud. These tools can help estimate the cost of deploying new applications as well. They can also help companies think through the dependencies and associated services that need to be enacted on a cloud environment and what it would cost on different cloud platforms.
Tools that rank usage and costs also support reporting capabilities as a service, allowing users to track trends in cloud use and estimate future costs. When a spike in costs or usage occurs, they can send alerts so that users don't end up with an unexpectedly high bill.
Examples of companies with these cloud cost analysis, monitoring and alerting tools include the following:
Committing to real benefits
On-demand pricing is generally the easiest cost model offered by cloud providers. It's the de facto standard, and 90% of the pricing models use some variation of on-demand pricing. These are similar across all providers and are charged usually per hour, although a few cloud providers are now offering per-minute pricing.
Longer-term contracts are a bit more complicated, but they give enterprise consumers the opportunity to reduce pricing with a certain level of commitment.
It is also more difficult to compare alternative pricing models across providers, in part because only 64% of IaaS providers surveyed publish their pricing online, said Rogers.
There is also still a fair bit of variation in the cloud market. If the cloud market is going to get more commoditized, Rogers expects pricing to get simpler. "Cloud providers want to show that they are cheaper, easier and quicker. The best way is with pricing that is easily comparable," he said.
Providers base their pricing on methods previously used in IT services. Rogers said that 45% of them offer a recurring method often used for managed hosting, and roughly one-third offer a pricing method similar to the "commit and burst" method used for bandwidth.
Mark Bakker, technical marketing manager at Copper.io, which makes the cloud pricing calculator and analytics service, observed that Google discounts are based on usage automatically with no upfront costs. This is suitable to the vast number of normal workloads that requires servers all the time.
But AWS only offers a discount to those who buy it up front. The AWS approach can be cost effective, but is unsuitable for long-lasting instance requirements.
He believes that Google's new pricing model is very disruptive as it's hard to see how AWS could respond in kind. Bakker said, "Price drops are one thing, but new models where there are a legacy of pre-payments is a whole other issue."
Looking at the big picture
The value pricing offered by basic IaaS services can be a bit misleading. Many enterprise consumers fail to understand the cost of doing price analysis themselves. Rogers explained, "There is lots of competition in the cloud market, and consumers are naturally looking at unit costs but not the additional cloud computing costs they will add on themselves, or how much these could be reduced from choosing a value-added provider."
Support is a classic example. If a company has had a relationship with a telco for five years, its IT staff members already know the people to call and can escalate a problem quicker. In theory, they should be able to resolve that problem more quickly. These options might not be available with a commoditized provider.
Initially, companies might consider the added value to be more expensive before considering the long-term cost of resolving an issue. At the moment, the market is melding between pure commodity and value-added providers. AWS, Azure and Google all provide some level of customer support.
It might also be useful to consider higher value services for applications with high security or privacy requirements that meet regulatory standards. Rogers expects to see community clouds for specific verticals that meet these more stringent requirements. For example, a PCI compliant cloud might come with on-demand pricing but at a higher price point that reflects the more rigorous requirements. This could also help with abstracting away the burden of keeping up with changing regulations.