Pretty much in anything we purchase, when given too many options, it can be exhausting to weigh up all the pros and cons of each offering. This is the case with the (supposedly) most liberating technology ever, cloud computing. When people select cloud services for their applications from cloud service providers, they rarely get the optimal solution. In fact, they almost never do, because the cloud service providers use an age-old sales strategy to provide so many options, typically in the hundreds, to make the purchase selection almost impossible to understand.

The act of securing units of IT on the cloud is a great example of the tyranny of choice. We all like to think we get the best solution for our organisation, one that will reduce our business risk, be cost-effective and increase our margins. However, the truth is that it is impossible to achieve, because there are too many variables to take into consideration.

Buying bundles of processing power, memory, storage and networking for your applications would be hard enough on its own. It’s even worse when you factor in the matching of your application needs to the right services. Which family of services should I be selecting? Within that family, which products do I select? Perhaps I need to be looking at containers? Do I optimise for memory? Is application “spiky” in nature, so therefore I need to be considering a service that matches that type of app? And if so, which one?! It is not humanly possible to consider all the variables, as each microscopic decision is a movable beast that is subject to change by the millisecond.

In response to the critique that service providers make their customers operate in the dark, the major cloud providers have devised offerings that purport to offer a solution. These are called Reserved Instances. Amazon Web Services (AWS), Microsoft Azure and Google Cloud responded to customer confusion over the various bundles they offered – by offering a new way to save money by making future commitments and buying in bulk sooner than you need the services. Have they helped? Not entirely, as they are almost as complicated as the system of tariffs they were meant to simplify.

Reserved Instances promise customers cost savings by committing to a certain level of consumption. However, buying Reserved Instances based on what is being used does little but to lock the organisation to making a long term purchase of the wrong mix of instances. Better to optimise first and only then realise the financial benefits Reserved Instances bring.

Dealing with that many variables – each with a huge leeway for change – you would be more likely to win the lottery than to get it spot on. While cost savings are always a priority for the purchaser, in actuality, it is risk that should be at the forefront of their mind. Mismatch between cloud provider resources and an organisation’s application demands can lead to serious business risk in addition to unnecessary spend. Is your application truly getting what it needs at the time it needs it?

There are tools to help you monitor change, but nothing to help you to keep pace with it. The most common mistake people make is to use a bill reader to analyse the cloud bill, but sadly, most of these management instruments are too blunt for a job that needs to be looking at the infrastructure issues first.

The only real way one can keep pace, is to use machines to track the machines, with deep multi-dimensional permutational analysis. This type of analysis and machine-learning ability can learn all the application workload dynamics, in order to work out the demand for resources.

The next step is to work out the supply side. Which isn’t easy, because the supply of processing power, storage and memory is another moveable beast, once again due to constant pricing changes and new technologies that are introduced regularly. Again, it is beyond the capacity of a human brain to keep up with them.

The good news is that next generation cloud optimisation solutions are here to help those who are trying to manage the pressures of application performance requirements, business risk and cost pressures. With machine-learning technology to make applications self-aware of their precise resource requirements – and if permitted – allow them to automatically optimise themselves 24/7, the complex mapping of supply and demand is achievable. Only then will you feel like you’re winning the cloud lottery on a daily basis.