Operations are going Distributed, Further Enabling Edge-computing

By Cornelia Davis, CTO, Weaveworks

One of the things that made the Internet successful while previous similarly distributed systems achieved only limited success was the highly distributed design of DNS. While conceptually a "single" lookup service mapping names to IP addresses, the fact that both the designs of the hierarchical naming system (domain names) and the DNS protocol (the data storage, update and lookup services) were widely distributed allowed it to operate in the constantly changing environment of the Internet. Of the utmost importance is that the operational model for DNS was designed to support local autonomy with global coordination.

As we have moved more and more of our computing to the cloud (private and public) the first area of focus for the industry has been around designing software that can execute well in the highly distributed dynamic environment that is the cloud - we've broadly adopted a set of cloud-native patterns for software design. But we've lagged in embrace and even an understanding of the operational patterns that work well in those same distributed environments.

2021 will be an inflection point. This is the year that we will begin to see common patterns emerge and be broadly adopted in the industry. IT systems will not only enjoy greater resilience, and security but the foundation will be set for massive growth and scalability. There are several signals that foreshadow this;

• Istio - The control plane allows for operators to manage centrally while policies are distributed to and enforced in the highly distributed service mesh. The embrace of Istio, Linkerd, AppMesh, the Open Service Mesh and  other service mesh technologies like Kuma and Kong Mesh are allowing us to test and refine these distributed operational patterns.

• Cluster API, which is seeing increased adoption with its growing list of providers, includes in its protocol a "move" command that moves the control plane for cluster management from a centralized bootstrapping service out to edge clusters.

• Git, a distributed system by design, is finding a whole new life in GitOps. GitOps is now a standard operating model for managing Kubernetes and its workloads on-premise, across clouds and at the edge.
  

The creation and refinement of distributed operational patterns could not be more timely with the coming of 5G. We simply have to have it there. For a good discussion on what GitOps and what a cloud native 5G looks like listen to our latest podcast:

https://www.weave.works/blog/kubernetes-at-deutsche-telekom-gitops-at-the-edge

The internet was expressly designed to be highly distributed. If DNS had been a centralized protocol, where all changes to the name service needed to be approved through a managed governance process, it would have failed. We need to do the same with the software that runs on that distributed substrate and cloud-native operating patterns like GitOps are key.

For more information on how you can implement GitOps in your organization, contact us. 

##

About the Author

As the Chief Technology Officer at Weaveworks, Cornelia Davis is responsible for the company's technology strategy so as to aid enterprises who are transforming their business though the leverage of cloud platforms. She cut her teeth in the space of modern application platforms at Pivotal where she was on the teams that brought Pivotal Cloud Foundry (Pivotal's PaaS), various data products and Pivotal Container Service (Pivotal's Kubernetes service) to market.  She is the author of the book Cloud Native Patterns: Designing Change-tolerant Software.