Application Security Groups, Deployments and YAML, Exams of Microsoft AZ-304, Microsoft AZ-304 Certifications, Network Security Groups

What to watch out for – Designing Compute Solutions

by Katie Campbell

In general, AKS and Kubernetes are more complicated than other technologies, especially Azure native alternatives such as App Service or Azure Functions. Additional tools are often required to better monitor and deploy solutions, which can sometimes lead to security concerns for some organizations. Although these can, of course, be satisfied, there is more work involved in setting up and using AKS for the first time.

Kubernetes is also designed to host multiple services and therefore may not be cost-effective for smaller, more straightforward applications such as a single, basic website. As an example, the recommended minimum number of nodes in a production AKS cluster is three nodes. In comparison, a resilient web app can be run on just a single node when using Azure App Service.

App Service

App Service is a fully managed hosting platform for web applications, RESTful APIs, mobile backend services, and background jobs.

App Service supports applications built in ASP.NET, ASP.NET Core, Java, Ruby, Node.js, PHP, and Python. Applications deployed to App Service are scalable, secure, and adhere to many industry-standard compliance standards.

App Service is linked to an App Service plan, which defines the amount of CPU and RAM available to your applications. You can also assign multiple app services to the same App Service plan to share resources.

For highly secure environments, Application Service Environments (ASEs) provide isolated environments built on top of VNets.

App Service is, therefore, best suited to web apps, RESTful APIs, and mobile backend apps. It can be easily scaled by defining CPU and RAM-based thresholds and are fully managed, so you do not need to worry about security patching or resilience within a region.

What to watch out for

Because App Service is always running, is always costs – that is, it is never idle. However, using automated scaling can at least ensure a minimal cost during low usage, and scale-out with additional instances in response to demand.