The use of the cloud is not governed by technology so much as it is governed by cost: the cost of on-premises management, support, expertise, and environment vs. the cost of cloud services and outsourced expertise, management, etc. The cost differential must be high enough in the short term to allow it to become valid in the long term. There are lots of cloud calculators out there. Since Apple, Dropbox, and others have changed clouds or moved to their own data centers, what does this tell us about the future of cloud?

The Future of Cloud

The future of cloud is still bright, shiny, and new. It will continue to grow, as is demonstrated by Azure, Amazon, Facebook, Apple, and other companies via their hardware purchases. They are spending quite a bit. However, the fact that Apple and Dropbox left Amazon is a blow to Amazon; there is no doubt about that. This has elicited several questions on social media:

  • How big do you need to be to build your own data center?
  • Is going to your own data center more about ego than reality?
  • Do the skills exist to use your own data center?

All of these questions theoretically deal with Infrastructure as a Service (IaaS) rather than Software as a Service (SaaS)–based cloud models. However, they all deal with the issue of Cost: cost of a data center, cost of the skills, etc.

Our Case Study

We here at The Virtualization Practice have used clouds for part of our infrastructure since the beginning. The choice to use each was based on two factors: cost and functionality. We have also chosen to use clouds based on our readers’ usage. We want to experience what our readers experience, and we have. We have also experienced spiraling costs and functional issues that have made the cloud less than desirable. Nevertheless, we still maintain several cloud presences. We first used the cloud to solve a problem, and we stayed in the cloud for convenience. Here is our adoption and implementation timeline:

  • We opened our file share & sync service to aid in sharing marketing material
  • We opened our customer relation management tool to aid in sales and marketing
  • We opened our newsletter management SaaS
  • We switched newsletter management SaaS providers based on both cost and functionality
  • We moved our customer-facing infrastructure to the cloud to solve a data center movement problem
  • After the data center moved, we pulled everything back on-premises except our website and associated databases
  • Cost issues drove us to look seriously at our CRM platform and then move from the cloud to an on-premises solution
  • Costs forced us to move our IaaS infrastructure for our website to a SaaS-based solution. After debating moving this on-premises, we decided instead to outsource the required expertise and use a much cheaper SaaS solution
  • Costs are forcing us once again to reconsider our newsletter management tools

As you can see, we still maintain several SaaS tools. However, we have moved entirely away from using IaaS within a cloud, due to cost and performance. To gain the performance and functionality we have now with SaaS by instead purchasing more IaaS instances would cost us triple, if not quadruple. Networking costs for use of on-premises installs would cost us 2-3 times our current networking bill.

It Is All About Cost

The future of cloud will be about cost. I foresee a time in the not-too-distant future when continuous integration and deployment will not only deploy to any cloud, but will pick the cloud into which to deploy based on daily, weekly, and monthly costs. I envision a time when ITOA will pull in not only log data, but also cost data and code-quality data to drive placement of code within various IaaS, PaaS, and SaaS solutions as required.
Currently, there are a few products that provide cost models for major cloud players; examples include Cloud Cruiser and Akasia. These use APIs to pull costing information and let you make predictions about where to place workloads. These types of tools need to be included in ITOA deployment decisions to make the cloud decision a simple one of cost. If the cost is too high, then report and try another cloud. Either way, the deployment has limits based on cost.

Closing Thoughts

This is the true utility model for cloud. Consider electricity for your home: you either have a single provider with choices of energy to use (such as wind, nuclear, coal, tide, etc.), a choice of providers, or the choice to produce your own electricity and sell it back to the grid—your home becomes part of the electric company for the purposes of power generation. Alternatively, you can use generators that do not sell back to the grid.
In other words, you either have on-premises power generation or cloud-like power generation, which almost everyone uses. Most who make a choice do so for political or cost reasons. Cost drives our bills, after all. Most people pick the cheapest option or the one that provides the most redundancy in case of power failure—thus, local sources make sense to use. Cloud needs to get to this point. Currently in the news, I see these costs appearing as an issue only for really big companies. However, cost drives these decisions for smaller companies, too. Yet, many smaller companies will choose to move between clouds such as we did—thus, from IaaS to SaaS.