Why Edge Orchestration is Different

STL Partners published an interesting report a couple of weeks ago titled “Why is now the time to rethink edge orchestration?” I fully agree with the entire contents of the report and here is my take on why edge orchestration is different. Different from what, you might ask? Different from telco MANO (management and orchestration) and cloud orchestration. 

I think there are four key reasons why edge orchestration is different:

  1. Scale: This is the most obvious reason. The scale of edge orchestration is orders of magnitude greater than traditional networks. It is not unusual to have tens of thousands of edge sites in the enterprise or telco context. Walmart announced 10K edge sites a couple of years ago. I was just talking to a senior executive at Telefonica, and he mentioned that there are 10K central offices and 20K baseband locations just in Spain. Orchestrating hundreds of network services or applications across tens of thousands of sites creates scale complexity.
  2. Heterogeneity: I compare the cloud to a departmental store. It is provided by one company; it is consistent, coherent, and orderly. The edge is a bazaar. Just at a cloud edge such as Equinix, you have multiple compute providers (e.g. Equinix Metal, AWS Outposts, HPE Greenlake, Zenlayer), multiple storage providers (e.g. NTAP, Dell APEX, Seagate Lyve, Pure Storage), multiple CaaS providers (e.g. Red Hat OpenShift, AWS EKS Anywhere, Google Anthos), multiple network function providers (Cisco, Fortinet, Palo Alto), and multiple connectivity providers (e.g. Equinix Fabric, Megaport, TATA Communications). Orchestrating environments in a bazaar is much harder than in a departmental store. 
  3. Dependencies: The whole mantra of the cloud has been to break the dependency between the workload and the underlying infrastructure. However, at the edge, this is impractical. Networking, machine learning, and AR/VR workloads have infrastructure dependencies. To further complicate matters, these applications also have network service dependencies such as Private 5G. In other words, at the edge, the infrastructure, applications, and network services all have to be orchestrated in a concerted fashion.
  4. Dynamism: The edge is dynamic. At the fall 2022 GTC, Chris Lamb from NVidia presented the concept of a multiservice RAN where, like Batman, the RAN infrastructure changes its entire persona from day to night. During the day (or periods of high utilization), the edge runs a RAN workload. During the night (or periods of low utilization), the edge runs AI/ML workloads. There are many other examples and reasons for why the edge is more dynamic than the cloud.

This means that existing orchestration mechanisms don’t work in this new paradigm, and that there is a need for a new set of tools. Please contact us to discuss this topic in more depth.

The Aarna Networks Multi Cluster Orchestration Platform (AMCOP) is purpose built for edge orchestration. Learn more here. Free consultations and trials are available.

We use cookies to enhance site navigation, analyze site usage, and assist in our marketing efforts. For more information, please see the Aarna Networks Cookie Policy.
Accept cookies