Cloud Computing looks to be making the transition from a development and test option to a primary considerations for many businesses, but many are finding the pricing models and options overwhelming and complicated to figure out.
In the fall of 2014, 451 Research released a report at their Hosting & Cloud Transformation Summit that included information about the top cloud computing-related pain points, which pricing/budget/cost rated as the number two pain point on that list behind security.
How does an organization determine the best cost for their environment, what should they buy, what cloud providers should they consider. Regardless of the fact that Google, AWS and Microsoft appear to be playing the pricing battle to try and win the title of lowest priced public cloud offering, cloud computing decisions are no longer based just on price. IT leaders are looking at a variety of criteria in their decision making process, and are looking for processes, tools and ideas to simplify this challenge for them.
Tale of Two Approaches
Our data shows that most organizations could be saving an average of 66% on their IaaS infrastructure by simply gathering performance data from their infrastructure and making decisions on application and workload usage vs. using an inventory approach (buying exactly what you have in the data center in the cloud). Then we have reserved capacity vs. on-demand capacity in AWS, which can cause some IT buyers a little anxiety.
We recommend for anyone moving to AWS for the first time or for a new application to purchase On-Demand instances to see how your application runs in AWS, and then if you are happy with the results than you can consider buying reserved instances.
Something to consider when buying reserved versus on-demand; you are committing to a 1 year or 3 year term when buying reserved instances, which can have a 30-50+% savings over on-demand pricing, but you are stuck with the hardware characteristics of that instance. What does this mean? When you buy reserved capacity you are buying specific technology that you will run your application for the next year or three years, with no upgrade options. In some cases, organizations have purchased reserved instances of a specific hardware type, and a month later AWS lowers the price on their newer hardware technology, but you are stuck at a higher price on older technology.
If your use case is suitable for that option, and it is the right fit for your business, then go with that option. But if your application is of dynamic nature and you need better performance on newer hardware technology throughout the year or over the next three years, than on-demand pricing may make the most sense for your organization.
Most of these issues are addressed specifically to AWS, but Microsoft has developed their model to align with AWS. Microsoft has pledged to stay comparable with AWS with respect to pricing. Do not assume that AWS or Microsoft are going to be the best option for price for your business, we see situations where AWS and Microsoft are not necessarily the best price for larger instances. In cloud we are not paying for what we use, we are paying for what we reserve, and in some cases we are paying too much.
Network I/O and Percentage of Daily Cost
We looked at 148 of our clients and looked at just network I/O, let’s talk about the percentage of daily cost. If you take network I/O and you are going to spend $10 a day on a cloud server, based on your network usage, what percentage of that cost is made up of network I/O?
Our analysis showed that most organizations saw less than 25% of their cost of cloud was made up of network I/O. In some cases the network I/O of a particular server can make up 50-75% of the total daily cost of a server running in cloud. In 2% of the cases we evaluated, organizations had servers/workloads that network would make up greater than 75% of the total cost of running that server in the cloud.
If you have large database servers with data replication and high network communication to those applications, the cost of running those applications in the cloud may be prohibitive. We believe that before you can truly determine if the cloud is the right place for your application, you should complete a thorough readiness assessment of each application and their dependent workloads. Know before you go!
Keys to Success with Cloud Pricing
How can you be prepared when moving to public infrastructure-as-a-Service, you need to make sure you do not overbuy from a capacity standpoint.
- General Usage information for CPU, Memory and Disk I/O – You need to have a clear picture of the workloads you plan to move to cloud, and this includes CPU, Memory, and Disk I/O to ensure you are not going to over pay capacity.
- Current network usage and impact on performance – Including network usage when planning cloud cost models is critical. In some cases, your servers network I/O may make up 75% of the total cost of moving to cloud. It would be great to know this in advance.
- Application dependency mapping – Prior to pricing out your cloud implementation, it is important to understand all of the associated dependent workloads that may have to be moved with your application. It is much easier to price an application stack than pricing individual workloads.
The great thing about IT today is that organizations have plenty of choices of where and how to run their applications. The problem is that with a ton of choices, it also makes selecting the best option for your business.
We recommend that prior to selecting a cloud provider, you should have a clear picture of your entire environment, including full application dependency visualization for each application and a full communication density map of all of your host and client connections. Don’t make the mistake that plenty of organizations have made before you, it could save you hundreds of thousands each year in unnecessary cloud cost.