Applications or parts of applications that operate near the edge are not the same as applications that thrive in the hyperscale cloud. In the same way that lifting and shifting a traditional application to the cloud doesn’t fundamentally change the nature of the app, and sometimes simply isn’t doable because of app requirements, you can’t simply shift cloud-native apps to the edge and expect good results. What’s different about edge-native apps?
Cloud-native apps are designed to leverage effectively unlimited horizontal scaling, with rapid change and rapid deployment as key. Innovation (at scale) is also central. Edge-native, on the other hand, is focused on real-time and dynamic automation of physical systems (things and people). Existing apps becoming cloud-native began as monolithic, back office apps. Existing apps that are becoming edge-native began as embedded computing, operational technology (OT) or independent end-user devices. Different apps have different fundamental requirements. Between traditional and cloud, the differences are about stability versus agility, vertical versus horizontal scale, monolithic versus microservices, etc. But between cloud-native and edge-native, a few things stand out as differences (at least at the extremes):
Hardware: Highly-standardized and abstracted hardware that lives in a data center, versus hardware variety, low abstraction, hostile environments and location awareness.
Resilience: Reliance on a cloud fabric that never fails, building resiliency to node and region failures into apps, versus an edge that is expected to fail, relying on the infrastructure architecture itself to manage resiliency.
Scalability: Horizontal and unlimited scaling, versus vertical and limited scaling, orchestration across edges and up the topology stack.
Elasticity: Rapid spin-up and -down versus limited elasticity.
Data: A process and store, centralized model versus caching, streaming, real-time and distributed model.
Orchestration: App self-orchestration in a horizontal way, versus infrastructure and edge orchestration across edges and hierarchically.
Security: A trusted fabric in physically secure data centers versus a trust-nothing environment in physically insecure locations.
App Model: Microservices and container-based, built for horizontal scaling, stateless, versus container- or VM-based, more monolithic nodes, often stateful.
Networking: Internet and rich capability versus varied and mobile networking.
Management: Centralized management, automation of software versus remote centralized management, fully automated hardware and software.
Of course, this is all on a spectrum. Those edge-oriented apps that live in a local datacenter or microdatacenter might have many more cloud-native attributes. But at the extremes, there are different requirements that require different technologies, processes, app architectures, and skills.