Last year was a landmark one for the service mesh. Linkerd attained CNCF graduated status, introduced traffic policy, and publicly powered companies like Microsoft and HEB. At KubeCon Los Angeles in November, we conducted a simple survey of attendees to understand the nature of their service mesh adoption.
We were curious because, at Buoyant, we had noticed a significant shift in the nature of service mesh adoption over the course of the year. Gone, or at least greatly diminished were the tech enthusiasts and neophiles excited about the technology for its own sake. In their place, the majority of Linkerd adoption was coming from engineers excited about the value of the service mesh, rather than the technology. These folks wanted a service mesh up and running as quickly as possible so that they could go back to the rest of their job.
To better understand this trend, we asked KubeCon attendees some basic questions. From the results, we uncovered a few key observations:
Read on for more!
We asked participants which feature of a service mesh mattered most to them. Mutual TLS was by far the top response, with 58% of respondents saying that the need for end-to-end encryption is driving their service mesh adoption. This came as no surprise: Linkerd’s zero configuration, on-by-default mTLS in 2020 has long been a driver of adoption for the project.
But we believe mutual TLS is just the beginning. Our conversations with platform and security teams last year have revealed a growing level of sophistication when it comes to cloud native security. While “just encrypt the traffic” is often the starting point, it’s followed immediately by demand for micro-segmentation and policy, with a strong desire for zero-trust implementations that require neither network-level nor host-level trust. The workload identity provided by mTLS is the foundation for this approach, and our Linkerd 2022 roadmap speaks exactly to where we believe the future of cloud native network security lies.
According to survey participants, the top factors slowing service mesh adoption were time investment and complexity, with more than a quarter of respondents referencing those factors. We’ve heard this time and time again: after adopting Kubernetes and all its machinery, there is rarely an appetite for additional layers of complexity to their deployments.
Here, Linkerd’s focus on simplicity and low operational overhead continues to pay dividends. In a space that’s notorious for complexity, Linkerd stands out as the simplest, most engineer-friendly solution, not to mention the fastest and lightest one. The project’s unique investment in modern proxy technology has allowed it to easily eclipse Envoy-based service meshes.
We asked attendees what they looked for when choosing between service mesh projects. The most frequent answer was ease of use and installation (25%); followed by the extent of feature-set (20%), simplicity (10%), and resource consumption (10%).
These answers echo those in the previous section: if time and complexity are the biggest blockers to adoption, then it makes sense to seek a service mesh that is easy to install and use. They also show that while some users haven’t adopted a service mesh because they believe it will add complexity and take time, others understand that there are service mesh projects that don’t necessarily lead to that outcome.
While service mesh adoption has matured, 63% of respondents reported that their service mesh deployments were not yet in production, while 14% said they have been in production for under six months. Given this was a KubeCon audience—presumably on the cutting edge of adoption—that means there is a lot of room for growth!
Not only is there plenty of headroom in existing Kubernetes adopters, the Kubernetes audience itself is growing! According to the Cloud Native Computing Foundation (CNCF), there are currently 5.6 million developers who use Kubernetes today, representing 31% of all backend developers and a 67% year-over-year increase. The future for Linkerd has never looked brighter.
Finally, we asked survey participants if they have switched service meshes. 28% of respondents reported that they had switched from one service mesh to another; of those, 80% switched from Istio to another service mesh.
While we didn’t ask participants why they switched (next survey!), responses to earlier questions already suggest some answers: complexity and ease of use are critical for not just adoption but for retention. We look forward to digging deeper into these questions in our next survey.
Our KubeCon Los Angeles survey produced interesting results that aligned with our anecdotal evidence of how and why Linkerd adoption was growing. Of course, this was hardly a scientific survey: the sample size was too small to confidently draw industry-wide conclusions; participation was limited to in-person KubeCon attendees who entered the vendor hall; and so on. That said, the results were largely in line with what we’ve seen and heard from Linkerd adopters in the real world, from the importance of mTLS for security to the struggle with service mesh complexity and the desire for simplicity.
For Linkerd, at least, the immediate future is clear: even more adoption, a continued investment in security, and above all else, a laser-like focus on our users and customers. As good as 2021 was for Linkerd, 2022 will be even better!