The on-demand and elastic nature of Azure

Azure is designed to scale. Workloads on Azure can easily scale up or down and scale out or in depending on demand. When coming from a data center mindset, where compute, storage, and networking capacity are preallocated and rarely changed, the cloud is a totally different ball game. Azure is all about pay-as-you-consume. If you don’t consume, you don’t have to pay. And this principle can be used to our advantage.

When you think of deploying a workload to Azure, start small. There is no need to pre-provision peak capacity when you can simply auto-scale based on metrics such as CPU, memory, or queue length. Also, consider shutting down development and test environment resources on weekends or holidays based on usage patterns. In this way, you can entirely avoid the cost associated with those resources.

All of this is possible only when you develop sensible KPIs and business metrics. Let’s look at how we can achieve...