Extreme Edge Computing: An Affordable Alternative Compute Supply

For the last decade, industry and academia have trailblazed the computing continuum. We've pushed computation from the centralized Cloud at the network's heart, through regional servers at the Near Edge, and out to 5G base stations at the Far Edge. This journey has enabled services that were once unimaginable.

But this trailblazing has consistently hit a wall at the network's final frontier: the Extreme Edge.

This is the domain of end-user devices, the true last mile of the network. It represents a vast, untapped reservoir of computational power hidden in the billions of devices we use every day. While traditional edge computing stops at the boundary of the enterprise network, XEC dares to go further, extending the continuum directly into the hands and homes of users. Our work formalizes this landscape, providing a systematic map to navigate its unique challenges and unlock its full potential.

That last mile, the extreme edge, is the most valuable real estate in the digital world. It offers two transformative advantages:

• Ultimate Proximity: You are right next to the end-user. This physical closeness is the only way to overcome the light-speed limitations of physics and deliver the microsecond latencies required for next-generation applications like true augmented reality, autonomous systems, and telesurgery.
• Massive Scale: You have crowds of powerful, connected mini-computers. They are sitting in our pockets, on our desks, and in our homes, closer than any enterprise server will ever be... often doing nothing.

So, why can't we just mesh them together and make them useful?

Unfortunately, it isn't that simple. To create a ubiquitous computing fabric out of this "troublesome mix," we must first understand and overcome its inherent challenges. Our research provides a comprehensive analytical framework that identifies 11 critical dimensions governing the design of systems at the extreme edge:

• Latency: The primary benefit, but requires real-time orchestration to achieve.
• Uncertainty: The core challenge, stemming from unpredictable human behavior and environmental factors.
• Resourcefulness: Devices are resource-constrained individually, but powerful when aggregated.
• Proximity: Enables low latency and privacy but requires sophisticated discovery mechanisms.
• Complexity: Orchestration is a highly complex, distributed, real-time problem.
• Modularity: Requires lightweight, platform-agnostic virtualization to deploy workloads.
• Connectivity: Relies on a heterogeneous mix of transient wireless links (Wi-Fi, 5G, D2D).
• Control: Service providers have no direct control over hardware, requiring indirect, influence-based strategies.
• Mobility: The network topology is constantly in flux as users and their devices move.
• Security: A zero-trust environment requiring robust sandboxing and validation to protect all parties.
• Privacy: Executing tasks on third-party devices introduces significant privacy concerns that must be technically and contractually addressed.

Today's convergence of communication and computation technology tells us that overcoming these challenges is no longer a dream. We believe that building a reliable computing fabric at the extreme edge has finally become possible.

The need for Extreme Edge Computing is driven by a powerful economic imperative: computing has become prohibitively expensive.

The chip market, which powers all high-performance computing, is suffering from a severe bottleneck. Unprecedented demand from the AI boom continues to outweigh supply, causing the prices of essential hardware like GPUs to hike to unsustainable levels.

This price surge is propagating through the entire supply and value chain. It creates massive capital expenditure (CAPEX) hurdles for network operators looking to build out their edge infrastructure and erects barriers to entry for startups, open-source communities, and academic researchers, hindering the pace of innovation.

This is not a temporary problem. We need an affordable, sustainable, and scalable alternative supply for computation. Extreme Edge Computing is a prime candidate to become that alternative.

Being a prime candidate is not enough. To unlock the potential of the extreme edge, we must tame its wild, unpredictable nature. This requires new models for the economic and efficient herding of both resources and workloads.

Defining the Landscape

We begin with a formal definition:

Extreme Edge Computing (XEC) refers to any form of computing that is done on non-dedicated hardware in the immediate vicinity of the end-user.

We identify two distinct "flavors" based on ownership, as this is the primary factor dictating the system's uncertainty:

• The User-flavored Extreme Edge: The crowd-powered flavor, akin to "Uber for computation." It harnesses idle resources on personal devices. Here, the primary source of uncertainty is unpredictable human behavior.
• The Enterprise-flavored Extreme Edge: A more controllable variant where devices are owned by a single entity but deployed in uncertain environments (e.g., a swarm of drones on a search-and-rescue mission). Here, uncertainty stems from the dynamic environment itself.

Making the Unpredictable, Predictable

The greatest challenge, particularly for the user-flavored edge, is reliability. How can you offer a commercial-grade service on an unreliable, crowd-powered substrate?

Our core contribution is developing analytical models that make this unpredictable environment predictable. We've created a "white-box" framework that mathematically links economic incentives to the performance of a device (like latency and throughput). This allows us to forecast capacity, manage resources efficiently, and provide the predictable service levels required for commercial applications.

Our Orchestration Toolkit

At the Queen's TRL, our toolkit includes a wide range of frameworks and techniques to orchestrate the extreme edge, including methods for:

• Splitting a workload and allocating it over heterogeneous, transient devices.
• Operating under uncertainty and with limited information.
• Identifying communities of devices to deploy workloads between them.
• Modeling and emulating resource usage and user behaviors.

Our exploration of Extreme Edge Computing is not just theoretical. We have validated our frameworks through major collaborative projects, advancing our work from foundational concepts to higher technology readiness levels.

The User-flavored Extreme Edge: The NSERC Alliance Project

We recently concluded a $3 million (CAD) project that ran from 2020 to 2024. In collaboration with an industry partner, this project deeply studied the user-flavored variant of XEC, allowing us to develop and refine our models against real-world challenges.

The Enterprise-flavored Extreme Edge: The DND Tactical Edge Project

We have recently launched a new project with Canada's Department of National Defence (DND). This work explores the enterprise-flavored variant, with a specific focus on powering semantic and effective communication at the Tactical Extreme Edge. This project tackles the unique challenges of deploying reliable computation in highly dynamic and mission-critical environments.

We are still in the process of exploring both flavors of XEC, and progressing from early technology readiness levels to widespread impact requires partnership. We are actively seeking collaborators to help us bridge the gap from foundational research to real-world deployment. We need:

• Support for Our Students: Funding and internship opportunities for our talented Masters and PhD students. Your mentorship and real-world experience are invaluable in shaping the next generation of innovators.
• Evolving Our Testbed: Access to equipment, AI-capable hardware, and advanced network emulation tools to evolve our testbed into a more mature, at-scale validation platform.
• Strategic Collaboration: Opportunities for field testing, access to real-world data, and strategic guidance from industry experts to ensure our research remains relevant and accelerates its path to market.