Balancing Innovation and Cost in the Cloud Era: Rethinking 'Best Practices' and the Role of On-Premises Solutions

For over the past five years, the default platform for hosting applications has been the cloud. This era has ushered in various kinds of technologies, people, and platforms, fueling what we could call the "cloud era." Terms like "cloud-native," "best practices," and various architectural paradigms have become commonplace in this landscape.

So, why this article? It's simple. Cloud technologies are game changers; they always will be. They have fundamentally changed how we design and interact with distributed systems. But for many businesses, there’s a bigger component that supersedes “best practices,” and that is cost.

Cost is a complex issue. Design principles state it should be the last thing to be checked after all design criteria have been met. But it also has to be the first thing considered because, without it, there’s no design.

Consider this: rent is affordable until you realize you can buy a house. Then, rent no longer seems cheap. Similarly, all costs are acceptable until they become unsustainable. The rise of cloud adoption has been fueled by generous cloud credits, propelling businesses to build and scale rapidly. But what happens when these credits run out?

I worked with a customer looking to reduce cloud costs. Half their bill was coming from RDS. Nothing fancy, just a DB with a replica. I asked, "Can this DB be moved to Kubernetes to eliminate these costs?" She responded, "This isn’t 'best practice.'" This is a common mistake—best practice is more than a reference architecture; it should be what is designed specifically for you. Better yet, you could call it “your practice.”

On-premises technologies didn’t evolve quickly enough, and this is partly why cloud technologies have skyrocketed. The various kinds of automation and the ability to compose and decompose as you like have been major factors. However, we’re entering a new era where many of these tools are becoming more universal. Many can be deployed on bare metal servers and will work perfectly fine. You can design your best practice for a set of bare metals. Yes, it isn’t multi-AZ or multi-regional, but it might be all you need.

On GCP, you can deploy disk volumes across regions and provision global load balancers that redirect traffic to multiple regions. But in truth, how many businesses need this level of high availability?

Consider these factors carefully when choosing a platform. If you can afford your systems outright, it might be worth considering. It’s even worse for businesses in developing nations, where they have fluctuating currencies and still need to remit the cost of their cloud platforms in dollars.

How do you plan your finances over the year? Tools are nice, best practices are nice, but before undergoing any technology project in today’s world, ask yourself: do I really need this?