Virtualization Technology News and Information
Article
RSS
Near, Far or Tiny: Defining and Managing Edge Computing in a Cloud Native World

By Keith Basil, VP of Product, Cloud Native Infrastructure, SUSE

Imagine managing a global deployment of 15,000 locations and 500 industrial Internet of Things (IIOT) devices at each site.  Your team has embraced containers and is 100 percent on board with a cloud native approach to meeting this challenge. So, where do you begin?  This scenario of 7.5M assets under management illustrates the core challenge of implementing and orchestrating cloud native approaches in the new, exciting, and yet potentially complex world of the edge.  

According to the Linux Foundation, "edge computing will be 4x larger than cloud and will generate 75 percent of data worldwide by 2025." Given the future size of the market and where we are today, we have quite a bit of runway in front of us.  Serious attention to the edge brings the realization that the law of large numbers is at play here, and we need to be ready to scale.

To add to the complexity, we see deep diversity in edge scenarios. From underwater deployments to satellites in space and everything in between, Kubernetes is being used to manage cloud native applications everywhere. Within the Kubernetes ecosystem, we have the facilities to tackle this. But before we dive into that, let's establish a framework for defining the edge.

Everyone has a different definition of the "edge." Collectively, we've found it useful to establish a baseline definition of the segments within the edge space so that meaningful discussions can ensue.

Defining the Edge

First, we must understand that we consider the hyperscalers, large data centers and core telecommunications infrastructure centralized infrastructure. We begin our definition of the edge from this perspective, as one of the fundamental first questions is "the edge of what?"

edge-near-far-tiny 

If we place the centralized infrastructure on the left side of our mental model, we then move away from centralized services to the right side of the world and into the edge segments. The first edge segment that we encounter is the realm of the communications service providers (CSPs).  Referring back to our mental model, we call this the near edge as it is nearest to the centralized services. In this edge segment, we find a diversity of use cases that address deploying compute and storage resources that meet the needs of the 5G core and multi-access edge computing (MEC) services as examples. One useful determinant that helps with our definition is to ask who owns and operates both the IP space and the infrastructure hardware in this segment. We find that CSPs own and operate both the IP space and the infrastructure hardware within the near edge. Multi-service operators (MSOs), such as a traditional cable company providing voice, video and data, fall into this category as well.

A logical and physical distinction also sharpens our mental map as the line of demarcation is extremely useful in crisply supporting edge segmentation where that IP space and infrastructure management is still, in most cases, managed by the CSP.  This line is important because there are near edge use cases that take the form of communications provider appliances that support next-generation services like software-defined wide area networks (SD-WAN) and Secure Access Service Edge (SASE). We see strong interest in Kubernetes at the core of these communications provider appliances. The implied notion of network and infrastructure ownership places these applications on the line of demarcation while remaining in the near edge segment as the communications providers own and manage those appliances.

Referring back to our mental model, the next segment we encounter is the far edge. Here, the IP space and infrastructure are typically owned, operated and managed by the end-user organization. It is the segment that is to the right of that line of demarcation.  Here we find a diverse set of use cases. Broadly, these fall into commercial, industrial or public sector deployments and can reach numbers into the tens of thousands of locations.  

Remaining true to our cloud native roots, the use cases in these industries are moving toward Kubernetes with varying cluster sizes. The main driver here is one of information and operational transformation where organizations desire to move cloud native applications to where they are needed to realize business value. We think most edge use cases will fall under the far edge segment, with data aggregation and analysis being the core function of the cloud native applications deployed here.  

This leads us to the final edge segment: the tiny edge.  This segment represents the world of fixed-function devices.  One of the main drivers here is the requirement to bring the IIOT under cloud native management.  These devices include sensors, actuators and IP cameras, which are typically found within the same layer 2 network as the Kubernetes cluster running in the far edge segment.  So, in essence, the tiny edge is a sub-segment within the far edge we've defined above.

The tiny edge is where the law of large numbers kicks in.  Capabilities in this segment are still maturing because of the diversity and quantity of IIOT protocols. In this segment, management solutions are sometimes proprietary and classically implemented with protocols that we must embrace if we are to bring these devices into our cloud native way of thinking.  We are encouraged by several upstream communities dedicated to enabling these gateway models and thus standardizing our approach to managing the devices in this space. 

Our Main Challenge

Given the law of large numbers and the diversity brought to the table, we see three pillars that are required to address the management at scale challenge we see at the edge.

First, we need a lightweight CNCF-certified Kubernetes offering.  Many of the far edge use cases require a single node cluster, which in deployment could be a system on chip (SoC) computer system, running a single or low number of application containers.  These are typically resource-constrained machines, so having a lightweight multi-architecture Kubernetes offering provides the best flexibility in meeting the needs of these deployments.  CNCF certification is vital as it standardizes our interfaces at the edge and allows us to leverage existing tooling and learnings.

Second, we need a lightweight, cloud native operating system that provides enhanced security due to its low attack surface and ease of lifecycle management. 

The third piece of the puzzle addresses the management at scale challenge head on.  We believe a GitOps approach to managing a large number of downstream clusters gives us the leverage we need to scale our team skill set. The Gitops approach is well tested with the cloud native world, and having a declarative source of truth addresses the complexity inherent in edge deployments quite nicely.

Collectively, we should work toward meeting these challenges.  SUSE will be doing this, and we ask that you join us in the areas we've outlined. We'd love to see the definition framework adopted, as we believe it enables us to have meaningful discussions that are efficient in guiding us to solutions that work.  Overall, we should strive to remove the complexity at the edge and focus on the business value that increases efficiency.

##

ABOUT THE AUTHOR

Keith Basil VP of Product, Cloud Native Infrastructure, SUSE 

Keith Basil 

Basil brings over 21 years of hands on experience in cloud and related industries. As Vice President of Cloud Native Infrastructure, Basil drives strategy and management of SUSE Rancher cloud-native products. Working with the SUSE global customer base, he is also driving development of cloud-native edge solutions that encompass cluster management, heterogeneous architectures, and zero-trust security approaches at scale.

Basil is also passionate about the next generation of decentralized cloud computing models. As an advocate in this area, he is working with communication service providers and public sector organizations to establish decentralized cloud infrastructure, applications and new revenue models.

Before Rancher, Basil led product management, positioning, and business strategy for security within Red Hat's Cloud Platforms business unit. Prior to Red Hat, he was instrumental in the design of a secure, high-performance cloud architecture that provided compute, storage and application hosting services for US public sector civilian agencies and contractors.

Basil holds a Bachelor of Science in Interdisciplinary Studies from Norfolk State University.
Published Tuesday, April 27, 2021 7:30 AM by David Marshall
Comments
There are no comments for this post.
To post a comment, you must be a registered user. Registration is free and easy! Sign up now!
top25
Calendar
<April 2021>
SuMoTuWeThFrSa
28293031123
45678910
11121314151617
18192021222324
2526272829301
2345678