Last week the Cloud Native Computing Foundation (CNCF) held KubeCon + CloudNativeCon North America. Even with conferences shifting from in-person to virtual, KubeCon still draws huge crowds and the entire industry's attention. Besides being one of the largest tech conferences of the year, KubeCon continues to show the cutting edge of technology at the forefront of the industry.

Toward the conclusion of the conference, Liz Rice - chairperson of the CNCF's Technical Oversight Committee (TOC) and VP of Open Source Engineering at Aqua Security - got on the virtual stage to share where the CNCF is going in the coming year and to talk about predictions for the industry as a whole. These predictions covered a vast landscape of new and emerging technologies and ideas. Some of the ideas are entirely within the bounds of the cloud native community, like service mesh, while others, like WebAssembly and eBPF, have even broader impact inside and outside of cloud native technology.

In the six years since the initial release of Kubernetes, the cloud native landscape has seen a proliferation of technologies and projects related to Kubernetes and cloud native in general. Rice even talks about this in her closing remarks, discussing the much loved and much talked about CNCF landscape. After adding many more graduated projects this year, one of the first predictions is that the coming year will see some current sandboxed projects at the CNCF fail. As Rice explains, this is a natural consequence of the CNCF pushing for innovation because not every innovative project will find a use case in the "real world" that justifies the effort of bringing it to market alongside juggernauts like Kubernetes, Envoy, and etcd.

CNCF's 2021 predictions

One of the most exciting segments was Rice's five predictions for the technology industry at large - inside and outside of cloud native technologies. These five technologies to watch (or six depending on how you count them) span several emerging technology platforms and speak to the great diversity of needs and projects in the open source community. The TOC's five technology trends to watch include:

  1. Chaos engineering
  2. Kubernetes for the edge
  3. Service mesh
  4. Web assembly and eBPF
  5. The developer and operator experience

Chaos engineering

The systems and applications we build are getting more and more complex and the human ability to accurately reason about how each component will interact and react becomes harder or impossible. Chaos engineering, first proposed and famously practiced by Netflix's engineering team, takes that change to heart and accepts that complex enough systems are genuinely unpredictable. Once you've understood this aspect of complex systems, the best way to test and reason about their reliability is to perform experiments that best represent real-life, unpredictable events.

While the concept of "turn off a component and see how the system as a whole reacts" makes sense on the surface, implementing such a methodology, especially in a large enterprise organization, can be daunting. Many projects and more than a few companies have been created to deal with this problem. It will be interesting to see if chaos engineering can move from the "elite" technology performers into a more mainstream engineering organization of every size and maturity level.

At GitLab, we have many customers already experimenting with or practicing chaos engineering. Uma Mukkara and Karthik Satchitanand from Maya Data presented on Chaos Engineering using GitLab templates and LitmusChaos at GitLab Commit in Brooklyn in 2019. We're also considering the many ways that chaos engineering could be more deeply integrated into GitLab as part of a single DevOps platform. Watch the video from Uma nad Karthik's GitLab Commit Brooklyn presentation below.

Kubernetes for the edge

Edge computing refers to an area of cloud computing where the infrastructure for computing, storage, and other requirements need to be placed in the field closer to users or their use cases. While cloud computing helps to centralize and create large data centers that benefit from scale, many if not most interactions with users occur far away from the data center and instead move to the edge.

As Kubernetes matures and transforms compute in the data center, more use cases for the core tenants of Kubernetes will emerge. And as those use cases expand in scope, we will continue to see new distributions or plugins to the Kubernetes ecosystem to support new use cases. Projects like KubeEdge, K3s, and others, bring the Kubernetes API and extensibility to more devices, even those on the edge.

With the onslaught of data, devices, and demand for performance, edge computing has become an essential component of many organizations' overall network topology. Bringing the flexibility and power of Kubernetes compute and processing options to this problem will continue to expand in the coming year. For example, there may even be a Kubernetes cluster running in your car today.

Service mesh

Rice predicts service mesh will be a hot topic in 2021, and with good reason. There has been an explosion of service mesh projects, discussions, and drama throughout the cloud native community in the past year. There has been an enormous proliferation of service mesh projects and teams discussing how a service mesh can benefit their deployments in 2020.

Similar to chaos engineering, service mesh attempts to organize the growing complexity of systems into a clear and reasonable package. As teams move to a microservices approach for application delivery, understanding the interaction and links between existing and new services becomes critical. Service mesh projects like Istio, Linkerd, and Consul have cropped up in the past few years. These tools help discover both known and new services and their connections. The goal of the projects is to create signal from noise, allowing humans to understand how those services interact and depend on one another.

In 2020, there was a lot of drama and discussion around the overall benefits and drawbacks of service mesh and the specific projects used to implement it. Now that there is a greater understanding among CNCF stakeholders about service mesh, we can expect the cloud native community to settle into a clear set of recommendations about when it is appropriate to implement a service mesh and how to make the right decisions about service mesh for your organization.

The most significant trend here will be with the ability of service mesh to not only discover services but secure them through policy enforcement. Additionally, the desire for observability will drive service meshes to become a critical cornerstone of observability in microservices environments.

Web assembly and eBPF

In this prediction, Rice rightly points out that the technologies of web assembly and eBPF are not - on the surface - related. Web assembly, also called Wasm, is a new type of virtual machine brought to the browser. eBPF is a programmable interface for interacting with the Linux kernel. So why did the TOC and Rice decide to include these two different technologies in one prediction?

Well, they share a common goal of sandboxing code when it runs. Sandboxing code, which means segmenting it from the parts of memory and the computer it doesn't need to get its job done, is a critical step toward allowing for secure code execution even of unknown sources. In the case of web assembly, that code is running in your browser. For eBPF, it could be running on a shared cloud-based Linux host. In both cases, these tools enable providers and security teams to effectively protect their code and data from prying eyes. This will remain a key objective for engineering teams for years to come, because we need to segment code better from a security perspective.

Securing code by segmenting processes

Many of the most massive zero-day attacks we've seen in the past few years demonstrate that some traditional pieces of the stack that we "take for granted" should instead be prioritized. Today, the barriers of the application memory or even CPU space are still ripe for attack. So inventing new and more secure ways of segmenting processes from one another will be a trend to watch for in 2021 and beyond.

At GitLab we see security and protection as belonging to the same DevOps lifecycle as the rest of engineering. The Secure and Protect stages of the DevOps lifecycle will continue to impact the rest of the cycle and how engineering departments develop and release code faster and more securely. We will see continued consolidation throughout the industry to bring security and protection initiatives to the forefront of every developer's mind, enabling developers and security professionals alike to deploy with confidence.

The developer and operator experience

Similar to prioritizing function over UX, our own experience in developing, deploying, and maintaining our projects often takes a back seat to "getting the job done." However, in much the same way, the developer experience and operator experience in their day-to-day tasks will be a key focus as technologies like Kubernetes enter a more mature phase.

We've already seen colossal consolidation and focus on the DevOps platform as a whole. It was just a year or two ago that we grudgingly accepted a disjointed set of poorly integrated tools, seeing it as unavoidable. Today, we see many DevOps companies and teams selling enterprise tools that are focusing on improving the dev and ops experience by building more capability into our devices and bringing together a more complete DevOps platform.

This is a mission that is obviously near and dear to our hearts at GitLab. Next year will bring a renewed focus on the dev and ops experience as more companies settle into the new normal of collaborating with teammates remotely, asynchronously, and automatically. This focus makes the DevOps platform we choose all the more critical to our engineering team's success, and as software defines the world we live in even more by the day, our organizations' overall success.

Developers and operators will come to expect an integrated DevOps platform that allows for the dual goals of getting software build and shipped on day 0 and maintaining and operating that software on days 1, 2, and beyond.

What's next?

A trend that is harder to quantify is the concept of observablity and growing trends toward more open communities. The concept of service mesh, Kubernetes at the edge, and the operator experience all play into observability, but I suspect we'll see more discussion of it in the coming year. Also the acceleration of 5G technology will impact all computing at the edge - Kubernetes or not. Beyond 2021, trends in AI in software development may accelerate changes to how we all interact. What trends do you think the CNCF missed in outlining things to watch in 2021? If you have a strong opinion, I'd love to hear about it on Twitter.

Git为Software Freedom Conservancy的注册商标,GitLab为GitLab B.V.的注册商标,我们已获授权使用“极狐GitLab”。