Leveraging Kubernetes to empower edge computing

[ServicesExperts]

Data is exploding at the edge. With this battle-tested orchestration platform and the right consulting partner, businesses can harness their data assets for a wide variety of use cases.

By Alex Tesch, Lead Consultant, Cloud Native Computing Practice, HPE Advisory & Professional Services

Kubernetes has become a de-facto standard for orchestrating cloud-native applications. The Cloud Native Computing Foundation (CNCF) recently announced the results of the 2021 Cloud Native Survey¹ and it is overwhelming to see that about 96% of organizations are using or evaluating Kubernetes.

Edge computing is another domain where more innovations are expected to happen in the next few years. It is expected that edge computing will be 4x larger than cloud and will generate 75 percent of data worldwide by 2025.

In this blog, we will discuss how Kubernetes can contribute to edge computing and the potential challenges it solves, and how HPE can help you on this journey.

Edge computing and its challenges

In simple terms, edge computing is all about moving compute power closer to users or applications so that the data can be processed in near-real-time. Going forward, some of the common use-cases of edge computing would be:

• Autonomous vehicles
• Predictive maintenance
• Multi access edge (MEC), virtualized radio networks and 5G (vRAN)​
• Smart homes
• Cloud gaming

… and more.

Depending on the use cases, edge computing can be classified into three different tiers: Near, Far and Tiny edge.

Near Edge: This is the segment closer to the centralized data center; the typical use cases include telco 5G/LTE private networks, cloud providers’ edge (point-of-presence, or PoP), etc.

Far Edge: This segment is farthest from the data center, and edge applications will reside on-site. It is most commonly used in the commercial and industrial sectors. Typical use cases include edge applications in retail stores, hospitals, oil and gas facilities, manufacturing units, factory floors etc.

Tiny Edge: This is the realm of fixed-function devices, which includes IoT devices. The key driver is to manage IoT devices just like any other cloud-native applications.

Now that we have discussed the use cases and categories of edge computing, let’s discuss the challenges it brings in as well. Some of the common challenges are:

• How do we efficiently manage and orchestrate these edge applications running across disparate environments (Near, Far and Tiny)?
• How do we tackle resource constraints in most of the edge locations?
• How do we make sure that security and compliance policies are consistent across all the clusters and subsequently reduce the attack surface?

We need an orchestration platform which is battle-tested to deal with all these challenges, and Kubernetes fits the bill perfectly.

Solving the edge computing challenges with Kubernetes

Kubernetes-as-a-Service is a new way of generating business value using Kubernetes, and most of the major tech vendors in the market have started switching to this model. Some of key features being offered are:

• Unified multi-cluster management – manage all Kubernetes clusters from a single pane of glass, whether it’s deployed in the cloud, on-premises or at the edge.
• Centralized application catalog – deploy applications to Kubernetes clusters easily in a few clicks.
• Consistent security policy and compliance management – role-based access control (RBAC) and cluster security/compliance policies can be applied to all the clusters using a GitOps approach.

Some of the products/solutions worth mentioning include:

HPE Ezmeral

RedHat OpenShift and its Multi-Cluster Manager

Suse Rancher

VMware Tanzu

Google Anthos

Microsoft Azure Kubernetes

Canonical Kubernetes

Most of the enterprise Kubernetes vendors currently support single-node Kubernetes cluster deployments, allowing Kubernetes to be deployed at edge locations with limited resource footprints. MicroK8S, K3S, K0S etc are some of the available options.

Adding GitOps to the equation will add more value; the usual DevOps CI/CD pipelines might not be a perfect fit when an application needs to be deployed across hundreds or thousands of clusters. Using the GitOps approach, application configuration can be treated just like application source code and the entire system is described declaratively. It also ensures that the approved changes can be automatically applied to the clusters and any drifts in the configurations can be identified and fixed automatically.

Another important point to be considered is the use of container-optimized host operating systems. These OS have tiny resource footprints, as they only employ packages with the most essential components that are required to run the container runtime. This can further reduce the attack surface. Examples of container-optimized OS include RancherOS, RedHat CoreOS, Flatcar Container Linux etc.

Kubernetes at the edge - Architecture

There are two architectural patterns primarily used while deploying Kubernetes at the edge.

1. All in one cluster. In this approach, the whole Kubernetes cluster (Control and worker node) is deployed within edge nodes. It is useful if the nodes have low resource capacity. SUSE K3S is a reference implementation of this approach.
2. Control plane in cloud. In this approach, Kubernetes Control plane will run in the cloud. Worker nodes will locate at the edge. KubeEdge is a reference implementation of this approach.

Another pressing question is, how do we manage the devices at the Tiny Edge like sensors, IoT devices, etc., using Kubernetes? Due to the low resource footprint, one might not be able to install a Kubernetes cluster on top of these devices. Project Akri is a CNCF project which tackles this problem by allowing you to easily expose heterogeneous leaf devices (such as IP cameras and USB devices) as resources in a Kubernetes cluster by leveraging and extending the Kubernetes device plugin framework.

How HPE can support your edge computing and Kubernetes journey

HPE’s line of servers like HPE SimpliVity, HPE ProLiant, and HPE Synergy, and intelligent storage solutions like HPE Nimble and HPE 3PAR, are used by numerous customers around the globe to run their Kubernetes-based application workloads, whether they’re in RedHat OpenShift, SUSE Rancher, Google Anthos, Canonical CDK, VMware Tanzu, Vanilla Kubernetes or other solutions. Furthermore, HPE Edgeline servers, like the HPE Edgeline EL300 shown below, deliver enterprise IT capabilities in a compact form factor designed for the difficult operating environments found at the edge.

Conclusion:  Cloud native computing transformation benefits not only the core platform, but also the edge – even more. 

HPE Advisory & Professional Services can help you get the most out of your edge Kubernetes design, implementation, migration, modernization, and operation throughout the transformation journey.  The Global Cloud Native Computing Practice within HPE Advisory & Professional Services drives the transformation with a proven methodology and vast experience/expertise.

To learn more, see our HPE Container Adoption Service solution brief.

Learn more about advisory and professional services from HPE Pointnext Services.

1. CNCF Annual Survey, 2021

Alex Tesch has been working with open-source enterprise technologies for the better part of a 20-year IT career in companies including Red Hat, IBM, Sun Microsystems, and now Hewlett Packard Enterprise. Alex is an APJ lead consultant in the hybrid IT Center of Excellence at HPE, where he evangelizes the open-source side of the cloud-native service portfolio and helps customers migrate legacy applications to container platforms. He enjoys running automation workshops, projects and seminars in the APJ region for cloud adopters.

Services Experts
Hewlett Packard Enterprise

twitter.com/HPE_Pointnext
linkedin.com/showcase/hpe-pointnext-services/
hpe.com/pointnext