If you weren’t at Kubecon SanDiego — Here is what you missed

Now that the dust has settled and the jet lag has finally worn off, it’s time to reflect on what was an awesome week at Kubecon San Diego.

Over the week we were treated to some amazing announcements, workshops, tutorials, demonstrations, keynotes, talks about all things CNCF, Open Source, and Kubernetes, and not to mention plenty of vendor swag! 😎

Once again, this Kubecon was the biggest ever, ~200 vendors, and with over 12,000 in attendance, it was almost double last years event in Seattle.

If a quick summary of announcements is what you’re after, Forbes: “10 Most Interesting Announcements From Kubecon + CloudNativeCon 2019” is a good read.

I want to share my key takeaways on various topics discussed at Kubecon, as well as general observations about the state of CNCF and Kubernetes.

• Kubernetes’ sprawling ecosystem
• Micromonolith, Monomicrolith, what?
• Kubernetes is somewhere between 10 and 45

Kubernetes’ sprawling ecosystem

Kubernetes is unlikely to have emerged as the de-facto standard for container orchestration if it hadn’t been for the overwhelming support of the CNCF, and a vibrant and growing open source community. In 2019, the total membership of the CNCF doubled with over 20 projects, 500 members, and 100 companies.

Unsurprisingly, the community has rallied to tackle what I would consider some of the most important or immediate of challenges, namely DevOps, Storage, and Monitoring and Observability, all of which are encompassed in the below chart.

In addition to the usual challenges of Kubernetes on traditional compute, it appears that the edge has become a new frontier. The edges continue to expand with Canonicals MicroK8s, and Rancher Labs K3s. K3s Under the Hood: Building a Product-grade Lightweight Kubernetes Distro gives a good overview of what K3s is and what it takes to build a lightweight K8s distribution for the edge. It’s not a fork of Kubernetes — remember that!

Meanwhile, there was an extremely strong presence of attendees and talks about using Kubernetes on both on-premise and public cloud; Whether it be hybrid on-prem and cloud, hybrid cloud, or all three. Kubernetes and its growing suite of tools presents itself as a compelling solution to ensuring portability, interoperability, avoiding vendor lock-in, and a consistent developer experience. It appears Google has caught wind of the trend and has begun heavily marketing the flagship Anthos.

Micromonolith, Monomicrolith, what?

The widespread adoption of Microservices has ushered in a new era in the industry.

However, solutions to old problems invariably create a new set of problems to solve.

Goutham Veeramachaneni & Edward Welch introduce the term “Micromonolith”, or “Monomicrolith” at their talk Cloud Native Architecture: Monoliths or Microservices?

They describe a common scenario in our industry where the manifestation of Conway’s law was easily managed by moving to Microservices, and then a subsequent growing feeling of nostalgia for the simple days of Monoliths.

Goutham & Edward’s talk offers some practical advice on application development that can in part bring back what was good about Monolith application development. He calls this paradigm “Monomicroliths”, linking to project Thanos, a highly available Prometheus implementation. Thanos allows you to deploy the tool either as a Monolith or as a set of Microservices. This ability greatly improves the developer experience on the project as it simplifies the local development environment. Other examples include Jaeger, Elastic, Grafana Loki, and Cortex.

This talk was very thought provoking for me and I think this is on the way to becoming an industry standard for large distributed system projects.

I am curious to see if Monoliths vs. Microservices is going to become a larger point of contention in the future. I think that the Cloud and Microservice community has long known that as we break our services up into smaller components, we increase the operational overhead and the difficulty of development.

The key point is that the ROI of this decoupling outweighs the cost of the overhead. How finely should we slice the pieces before they become useless?

Even the most staunch of MicroService advocates would know, there is an upper limit.

The community has in some part reacted to this issue, from API Gateways, and Service Meshes, to Minikube and even plain old docker-compose.

Time for questions was rarely had, and in the talks that I attended when there was, questions were sparse. This talk had probably the most questions I had seen, perhaps speaking to how thought provoking it was. To some, perhaps questioning the dogmatic belief that MicroServices are holy!

In the “Kubernetes at Reddit talk” Greg Taylor brings up “everyone’s second favourite topic that is not service meshes ….Local development environments”.

Greg asks the crowd if everyone is happy with their local Kubernetes developer environment, to which a member of the crowd answers with a rather loud and aggressive “NO!“. One member of the audience did have their hand up, to which the other crowd member responds “He’s a liar!“, and Greg responds “Yeah probably”. This was a pretty humble moment in the conference and I’m happy it was captured on video.

Kubernetes is somewhere between 10 and 45

Kubecon is a great networking and socialising opportunity for people in the industry. It’s particularly welcoming to newcomers with as much as 73% being first timers to KubeCon + CloudNativeCon North America in 2018 (and I imagine similar in 2019). In addition to this I found that many were very much in a discovery phase or completely new to Kubernetes.

The job market for Kubernetes talent is remarkably strong, with the CNCF announcing the CNCF Job Board, and according to Indeed.com 12,503 jobs globally which is up 9,934 from 6 months ago, there’s a certain feeling that Kubernetes is a strong, mature, and proven field, and one can seldom argue with the market.

Internally though, maybe it doesn’t yet feel that way, in Bryan Liles keynote “In Search of the Kubernetes “Rails” Moment” he stresses how Kubernetes needs a cultural and technical transformation which can sufficiently reduce its complexity. Stressing easier use, good defaults, self-service configuration, etc. It’s hard to argue with this, and I feel that this maturation step needs to happen, but when we put it into perspective with how far the project has come, it is a welcoming challenge.

If you would like to see more Kubecon content, most of the sessions have been uploaded to Youtube on the CNCF page, as well as the accompanying slides on the Sched website under the appropriate event.

Otherwise, to learn more or connect with us, head on to Momenton, and hit us up for a chat. We’re hiring! 😃

That’s all for now, stay frosty 😉