As we’ve seen, several vendor-specific CaaS services are available on the market. Still, the problem with most of them is that they are tied up to a single cloud provider. Our container deployment specification then becomes vendor-specific and results in vendor lock-in. As modern DevOps engineers, we must ensure that the proposed solution best fits the architecture’s needs, and avoiding vendor lock-in is one of the most important requirements.
However, Kubernetes in itself is not serverless. You must have infrastructure defined, and long-running services should have at least a single instance running at a particular time. This makes managing microservices applications a pain and resource-intensive.
But wait! We said that microservices help optimize infrastructure consumption. Yes—that’s correct, but they do so within the container space. Imagine that you have a shared cluster of VMs where parts of the application scale with traffic, and each part of the application has its peaks and troughs. Doing this will save a lot of infrastructure by performing this simple multi-tenancy.
However, it also means that you must have at least one instance of each microservice running every time—even if there is zero traffic! Well, that’s not the best utilization we have. How about creating instances when you get the first hit and not having any when you don’t have traffic? This would save a lot of resources, especially when things are silent. You can have hundreds of microservices making up the application that would not have any instances during an idle period. If you combine it with a managed service that runs Kubernetes and then autoscale your VM instances with traffic, you can have minimal instances during the silent period.
There have been attempts within the open source and cloud-native space to develop an open source, vendor-agnostic, serverless framework for containers. We have Knative for this, which the Cloud Native Computing Foundation (CNCF) has adopted.
Tip
The Cloud Run service uses Knative behind the scenes. So, if you use Google Cloud, you can use Cloud Run to use a fully managed serverless offering.
To understand how Knative works, let’s look at the Knative architecture.
