More analysts, pundits make wrongheaded argument about edge ‘eating’ cloud

What they said

This statement from an analyst does not make a lot of sense: Latency is the big killer of cloud computing, especially public cloud computing,” while an executive opined that “Orchestrated edge systems will become a viable public cloud alternative.”

The key words are ‘killer’ and ‘alternative’, which suggests that edge will somehow replace cloud. But reducing latency is something that the public cloud can and will address. It will just do it differently, in smaller increments, and in varied deployment models. Cloud computing regions are built at scale, with large ‘core’ availability zones spread across a given metro location. Should the latency of a given workload be an issue, public clouds have already built a number of solutions for customers. They can deploy on-premise with an appliance (compatible with the public cloud API) that connects to the public cloud, while the likes of AWS and Microsoft Azure are starting to build small blocks of compute in wireless and mobile connectivity centers.

In markets where no core region is present, public cloud providers are standing up edge zones. A big part of the value proposition around these edge-deployment scenarios is that the compute infrastructure addresses an edge workload requirement but is connected back and integrated with the core cloud region. Compute infrastructure is at the edge and not the core, but it is working in tandem with the core and driving growth there (the core) through uptake in various services ranging from analytics, archiving, long-term storage, cost optimization, database processing and more. It is simply not the case where edge computing displaces cloud computing. Instead, it is more of an extension of the cloud and is working in tandem with it. If that is how edge works, it will be difficult to envision how we are on the cusp of something boldly new and different.

The (wrong) view of edge data processing

Another view expressed from the pundit community goes something like this: “Edge computing provides the ability to run near-real-time data collection, analytics, decision-making and execution.” Data collection and certain functions may happen in something like real-time at the point of interaction at the edge. But most of the processing, analytics, crunching, application execution and storage is going to happen back in the core and not at the edge. They will not be separated and exist on separate and isolated islands, cut off from each other, which is what this argument implies. Edge computing is too small to handle so many different functions and it would be difficult to get scale and cost efficiency. Many application and analytics functions are still going to happen at the core.

Why the cloud versus edge debate is important

This is not just a conceptual or definitional debate. Understanding how edge and core compute are going to work together will define how the industry grows and make for correct modeling and growth projections. If some pundits are to be believed, edge scenarios will displace core workloads and eventually slow growth in the cloud, with edge emerging as a new category. But the reality is that is not going to happen. Pushing out to the edge is going to be an extension of existing architectures — basically pushing the application front-end out of the core — and any cannibalization of raw computing growth will be mostly offset by growth in storage, core compute, along with tools and services usage. The edge will have a certain ‘rising tide lifts all boats’ effect. It is far from cannibalization and replacement, which is what anyone trying to advance the ‘edge eats cloud’ narrative is mistakenly predicting.

About the author

Phil Shih is managing director and founder of Structure Research, an independent research firm focused on the cloud, hosting and data center infrastructure service provider markets.

Article Topics

AWS  |  Azure  |  cloud infrastructure  |  edge cloud  |  enterprise architecture  |  market research