[Paper Review] An Online Framework for Ephemeral Edge Computing in the Internet of Things

Gilsoo Lee, Walid Saad+1 more|arXiv (Cornell University)|Apr 18, 2020

IoT and Edge/Fog Computing4 citations

TL;DR

This paper proposes an online framework for ephemeral edge computing in IoT, where mobile edge nodes dynamically offload tasks to neighboring nodes in real time with unknown future task arrivals. Using a primal-dual online greedy algorithm, the framework jointly optimizes communication and computation latency, achieving near-optimal task allocation with a competitive ratio guarantee and an optimality gap of no more than 7.1% compared to the offline optimal solution.

ABSTRACT

In the Internet of Things (IoT) environment, edge computing can be initiated at anytime and anywhere. However, in an IoT, edge computing sessions are often ephemeral, i.e., they last for a short period of time and can often be discontinued once the current application usage is completed or the edge devices leave the system due to factors such as mobility. Therefore, in this paper, the problem of ephemeral edge computing in an IoT is studied by considering scenarios in which edge computing operates within a limited time period. To this end, a novel online framework is proposed in which a source edge node offloads its computing tasks from sensors within an area to neighboring edge nodes for distributed task computing, within the limited period of time of an ephemeral edge computing system. The online nature of the framework allows the edge nodes to optimize their task allocation and decide on which neighbors to use for task processing, even when the tasks are revealed to the source edge node in an online manner, and the information on future task arrivals is unknown. The proposed framework essentially maximizes the number of computed tasks by jointly considering the communication and computation latency. To solve the problem, an online greedy algorithm is proposed and solved by using the primal-dual approach. Since the primal problem provides an upper bound of the original dual problem, the competitive ratio of the online approach is analytically derived as a function of the task sizes and the data rates of the edge nodes. Simulation results show that the proposed online algorithm can achieve a near-optimal task allocation with an optimality gap that is no higher than 7.1% compared to the offline, optimal solution with complete knowledge of all tasks.

Motivation & Objective

• To address the challenge of transient, short-lived edge computing sessions in mobile IoT environments where tasks arrive unpredictably and devices frequently leave the network.
• To design a real-time, online task allocation framework that operates without prior knowledge of future task arrivals or system state.
• To jointly minimize communication and computation latency in ephemeral edge computing by dynamically selecting optimal neighboring edge nodes for task offloading.
• To develop a competitive online algorithm with theoretical performance guarantees using the primal-dual method, ensuring near-optimal task completion under time constraints.

Proposed method

• The framework models task offloading as an online optimization problem where tasks are assigned to neighboring edge nodes in real time, based on current knowledge only.
• A primal-dual online algorithm is designed to dynamically allocate tasks while maintaining a competitive ratio bound derived from the problem’s structure.
• The algorithm uses a dual variable update rule based on task size, data rate, and computation capacity, with a parameterized upper bound on resource usage via the variable $ z_j $.
• The competitive ratio is analytically derived as a function of task sizes and edge node data rates, ensuring performance bounds even under uncertainty.
• The method incorporates a time-constrained objective function that balances task completion rate and resource utilization across edge nodes.
• The framework ensures feasibility by enforcing constraints on total task load per edge node using a logarithmic barrier function and a parameter $ c = (1 + \delta)^{1/\delta} $.

Experimental results

Research questions

• RQ1 How can edge nodes efficiently allocate transient computing tasks in real time when future task arrivals are unknown?
• RQ2 What is the theoretical performance limit of an online task offloading algorithm in ephemeral edge computing environments?
• RQ3 How can communication and computation latencies be jointly minimized in a dynamic, mobile IoT setting with no future knowledge?
• RQ4 What is the competitive ratio of an online algorithm in this setting, and how does it depend on system parameters like task size and data rate?
• RQ5 Can the proposed online algorithm achieve near-optimal performance compared to an offline optimal solution with full knowledge of all tasks?

Key findings

• The proposed online greedy algorithm achieves an optimality gap of no more than 7.1% compared to the offline optimal solution, demonstrating strong practical performance.
• The competitive ratio of the online algorithm is analytically derived as a function of task sizes and edge node data rates, providing theoretical performance guarantees.
• The primal-dual approach ensures that the algorithm maintains feasibility while dynamically adapting to incoming tasks and changing network conditions.
• Simulation results confirm that the algorithm effectively balances load across edge nodes and maximizes the number of completed tasks within the ephemeral time window.
• The framework successfully handles the dynamic nature of mobile IoT environments by enabling real-time, decentralized task offloading decisions with minimal latency.
• The use of the parameterized dual variable update with $ c = (1 + \delta)^{1/\delta} $ ensures bounded growth of resource usage and enables tight performance bounds.