As AI continues to expand its presence at the Edge, Nvidia Jetson has emerged as the preferred platform for solution providers. Its GPU capabilities, combined with lower costs, have led to widespread deployment, especially given the extensive network of OEM Nvidia partners who create customized hardware for various use cases.

In recent years, Jetson has been utilized in numerous Edge AI applications, such as intelligent traffic management, retail analytics, theft detection, and smart farming. Due to its advanced capabilities and affordability, Jetson devices are being implemented on a large scale across thousands of locations. However, this widespread deployment is repeating a critical mistake reminiscent of past OnPrem deployments. Historically, companies requiring local compute power utilized diverse solutions from multiple vendors, each introducing its own device that had to be individually installed, configured, and integrated into the customer’s network.

Consider a well-known fast-food franchise running various applications in their stores, including kiosks, video surveillance, back-office operations, point-of-sale systems, and security. Each application came from a different vendor, each deploying its own device to run a single application. This approach is problematic, leading to enormous CAPEX for the customer and resulting in long-term inefficiencies. Adding a new solution necessitates installing an additional device, while removing an application means discarding the associated hardware. Moreover, many application vendors provided solutions as embedded systems, tightly coupled with the hardware. This was common in traffic management and IoT solutions, where the device was deployed and then largely forgotten until it was replaced.

This inefficiency was once acceptable due to several factors:

• ● Basic solutions with minimal compute requirements and low device costs.
• ● Simple applications that required little to no updates or maintenance.

Today, solutions have evolved, offering advanced features and heavily relying on AI to deliver innovative capabilities. These advancements necessitate powerful GPUs in small appliances that can be deployed anywhere, a strength of Nvidia Jetson. Unfortunately, the industry is repeating the same mistake, following the embedded system path. Currently, each AI application is supplied with a Jetson device for a specific use case. If a customer wants another solution, they purchase it from another vendor with another device. These solutions, following the embedded tradition, offer limited or no maintenance and updates, treating the device as a CAPEX investment intended to last its entire lifecycle.

Nvidia Jetson, a Cloud Native story

As Nvidia Jetson devices continue to advance in AI computing capabilities—and with their costs rising—it no longer makes sense to view them as mere embedded devices tightly coupled with the applications they were originally provided with. Instead, they should be seen as versatile platforms capable of running any application from any vendor, with the flexibility to replace or update those applications as needed. This aligns perfectly with the goals of cloud-native development, which aims to enable application deployment anywhere, independent of the underlying operating system. Today’s Jetson Edge devices are powerful enough to run multiple applications simultaneously, further underscoring the importance of decoupling applications from the hardware.

Stacked Edge vs Converged Cloud Native Edge

To remain competitive, companies deploying AI solutions must continuously offer updates and new use cases to their customers. For example, an AI Vision analytics application for the retail industry might regularly update its models for improved accuracy or to add new detection features. More broadly, service providers, managed service providers (MSPs), and system integrators (SIs) are always seeking new revenue streams and ways to maintain ongoing relationships with their customers. This can be achieved by offering new AI applications that can be seamlessly deployed on an existing Edge infrastructure.

Achieving this vision requires a robust underlying infrastructure and a strong application orchestration framework that spans all Edge locations. Such a framework should allow for the deployment, updating, and deletion of applications on the fly, without the need for hardware replacement or individual operations on each device.

Kubernetes naturally comes to mind when discussing cloud-native orchestration, as it has become the leading framework for deploying and managing hundreds of applications across multiple locations. It enables seamless deployment, updates, and management of containerized applications while offering self-healing capabilities, automated rollouts, and efficient resource management to ensure applications remain available and performant, all with minimal manual intervention.

Challenges of Cloud Native Edge

It's evident that deploying AI applications at the Edge and scaling them to thousands of Jetson devices requires a robust infrastructure and orchestration framework While Kubernetes stands out as the best option, it remains a relatively complex solution to implement and manage. In the cloud, providers like Azure, AWS, and GCP have mitigated this complexity by offering managed Kubernetes services, which simplify the process of setting up and managing clusters for their customers. Unfortunately, no equivalent solution exists for the Edge.

As a result, companies looking to leverage Kubernetes for their Edge orchestration must possess the necessary expertise and ensure their teams are adequately staffed. While this may be feasible for large corporations that already have Kubernetes engineers on board, it's a significant challenge for today's Managed Service Providers (MSPs), Systems Integrators (SIs), and other smaller organizations that may lack such specialized resources.

Namla, a Cloud Native Edge platform for Jetson

At Namla, we are committed to scaling Edge AI across thousands of locations, recognizing that such scalability requires the agility and performance achievable only through a Cloud-Native Architecture. Kubernetes stands out as the optimal choice for this purpose.

By utilizing Kubernetes as our core orchestration platform, Namla provides a robust cloud-native environment that simplifies the deployment of containerized and VM-based applications. Our solution eliminates the complexities associated with setting up and maintaining a Kubernetes cluster, enabling effortless application deployment.

Namla has also pioneered the first Cloud-Native SD-WAN solution, fully integrated with Kubernetes. This ensures secure and reliable edge-to-cloud connectivity while delivering advanced networking capabilities. The integration streamlines the management of distributed Edge infrastructures, enhancing both performance and security.

In addition, Namla offers comprehensive remote management and monitoring features, providing deep visibility into the entire infrastructure. This enables swift remote interventions, reducing troubleshooting time and minimizing downtime, ensuring continuous and efficient operations.