Why Automated Azure Cost Tools Aren’t Enough

Are you leaving money on the table (or, rather, in Microsoft's pockets) by relying only on automated "cost optimisation tools" to reduce your Azure bill?

There are a plethora of automated tools available these days, which will scan your Azure setup and make recommendations for cost savings. A lot of these tools are very good and we use them as well. They just don't tell the whole story, especially if you are developing applications on Azure - so you might still be paying a lot more for Azure than you need to, even after using an automated tool.

The automated tools are generally very good at identifying where you can save money using reserved instances, hybrid licensing, savings plans and so on. The good ones will also analyse your actual usage of resources vs the provisioned throughput, thus helping you to right-size and even find unused resources. This can yield substantial savings and is an excellent first step.

But there are things that an automated tool just can't do, especially when you are developing your own applications on Azure. At NewOrbit we have developed applications on Azure for 15 years and we have learned a few things. These are experiences we share with our Azure customers to help them further reduce their bills.

Some examples include:

• Are you using the best Compute hosting platform for your specific application? There are many ways to host your application on Azure and Virtual Machines are rarely the best choice. However, knowing which compute service to use depends heavily on the nature of your application and the tech stack it is built on. Platform-as-a-Service and Azure Container Apps are often (but not always) a better choice and generally offer better dynamic scaling options, which tends to reduce cost.
• Using Queues in your application can reduce hosting costs by reducing the minimum compute level that must always be available. I.e. queues allows you to scale both the application and often the database down lower than you would otherwise be able to, which translates directly to cost savings.
• Is the "hot" or "cold" storage tier optimal for your application, given how your application uses storage? And should you use a lifetime management policy to change storage tiers for some data over time?
• Can you split your data between different database and storage products in Azure? In some cases, this can unlock 99% cost savings for some data (yes, really).
• Which of the half-a-dozen-plus ways of hosting SQL in Azure is most cost-effective given your application requirements?
• Are there particular aspects of your code that is expensive? For example, "chatty" applications are much more expensive in a cloud-environment compared to on-premise. Using tools such as AppInsight you can spot this and often it is a very simple code change to significantly reduce the "chattiness".

As you can hopefully see from the examples above, understanding your application unlocks a range of cost optimisation strategies that a generic tool simply won't be able to identify.

• See Designing for scale in the cloud 101 to learn more about using queues and different database types.
• See Spend less on Azure for a much more in-depth guide to how to find costs savings.

Or ask us to have a look: We have already helped many other software development organisations reduce their Azure bills. See more and get in touch for an Azure Cost Review.