2020 has broken many rules about software development. Some perceived trends, however, remain myths

Some trends come and go, some trends stick. I remember when Docker launched and it was barely mature enough to run unit tests, and now containers are a common way for modern orgs to deploy applications. We’ve seen the rise of microservices but also the return to the monolith. We’ve seen “observability” take off in popularity, while serverless remains a buzzword with limited (but fast-growing!) adoption. Machine learning and artificial intelligence are still click-worthy headlines, but how much actual ML is being used within our applications?

The True State of Software Development

Let’s look at five common myths about the current state of software development and evaluate why they’re not completely true:

Digital Transformation Initiatives Will Be Hindered by the Pandemic

We live in an online world where every business is already a software business (or rapidly becoming one). The impact of the pandemic further highlighted ISG’s research which stated that “2020 will still see the rapid scaling of digital initiatives across industries.” While some industries have been hit so hard that digital projects had no choice but to stop (e.g. the airline industry), in many circumstances this will be the catalyst to fully embrace digital transformation.

The Waterfall Method Still Works

Organizations that haven’t adopted more agile and iterative approaches to building software are operating at a significant disadvantage. I know there are many big companies that still believe in doing all of the strategic planning upfront (I know this because I talk to these companies), but the truth is that it leaves very little room for flexibility and improvement in real-time. In today’s world, the customers and the users are kings—it is no longer possible to build great products without their input. If you take anything away from this article, it’s that the waterfall methodology needs to be abandoned once and for all.

Everyone Loves Kubernetes

K8s is not a one-size-fits-all solution. The fact of the matter is that a large chunk of the enterprise still hasn’t fully embraced the cloud, let alone started containerizing their applications. So while we have to give credit where it’s due and acknowledge that K8s is the clear leader for orchestrating containers, we also have to acknowledge that the hype doesn’t quite match the adoption. It’s still a relatively young technology.

According to a report from Datadog last year, 45% of their customers that run containers use Kubernetes. That number is growing and is certainly impressive, but when you realize that only a chunk of their customers run containers in the first place and they are a modern monitoring solution where we’d expect to see a higher percentage than a random sampling, it’s clear that K8s is far from mainstream.

Everyone Is Doing Tracing

The rise of the digital age has put monitoring and observability in the forefront. As we rely more and more on online services for internal and external interactions, we better make sure they are working properly. But at the same time, regardless of industry headlines and conference talks focusing on tracing, the truth is that the industry at large has yet to adopt it.

For most organizations, there’s so much room for improvement in getting logging and metrics right, that there’s no need to opt for tracing (yet). After all, Google built most of its global platform relying on metrics. Unlike SREs, developers often find tracing alien and very cumbersome to deploy and wield. Instead, more and more companies are opting to empower their engineers to see what’s going on in remote environments through software understandability.

AI Will Solve Our IT Problems

We’ve been hearing about the promises of AI in IT for a long time now. One of the terms still actively promoted that I am very skeptical of is “AIOps.” What does that even mean, really? According to a recent report from OverOps, only a very small percentage of companies are even considering AIOps in their plans for the future.

That said, I do believe in making better, data-driven decisions. So the more data we can collect, contextualize and understand, the better. We see how this helps companies identify anomalies, alert on issues and even project out into the future. But none of this, really, is artificial intelligence.

Conclusion

There’s often a gap between hype and actual adoption. That’s the role of good marketing. And ultimately, hype helps drive adoption. But as technical practitioners, we need to stay rooted in reality in software development and beyond. So while I am convinced agile has proven to be the best way to build and deploy software, and that most organizations would benefit from letting go of old-fashioned waterfall processes; on the other end, I haven’t been fully convinced of the benefits of tracing and AIOps for most companies.