[Paper Review] An Online Secretary Framework for Fog Network Formation with Minimal Latency

Gilsoo Lee, Walid Saad+1 more|arXiv (Cornell University)|Feb 18, 2017

IoT and Edge/Fog Computing9 references9 citations

TL;DR

This paper proposes an online secretary framework for dynamic fog network formation in hybrid fog-cloud architectures, enabling a fog node to select optimal neighboring nodes for task offloading under uncertainty of future arrivals. Using an exploration-exploitation algorithm inspired by the online k-secretary problem, the framework minimizes maximum computational latency and achieves near-optimal performance—within 2.7% of the offline optimal solution—while effectively distributing tasks between fog nodes and the cloud.

ABSTRACT

Fog computing is seen as a promising approach to perform distributed, low-latency computation for supporting Internet of Things applications. However, due to the unpredictable arrival of available neighboring fog nodes, the dynamic formation of a fog network can be challenging. In essence, a given fog node must smartly select the set of neighboring fog nodes that can provide low-latency computations. In this paper, this problem of fog network formation and task distribution is studied considering a hybrid cloud-fog architecture. The goal of the proposed framework is to minimize the maximum computational latency by enabling a given fog node to form a suitable fog network, under uncertainty on the arrival process of neighboring fog nodes. To solve this problem, a novel approach based on the online secretary framework is proposed. To find the desired set of neighboring fog nodes, an online algorithm is developed to enable a task initiating fog node to decide on which other nodes can be used as part of its fog network, to offload computational tasks, without knowing any prior information on the future arrivals of those other nodes. Simulation results show that the proposed online algorithm can successfully select an optimal set of neighboring fog nodes while achieving a latency that is as small as the one resulting from an ideal, offline scheme that has complete knowledge of the system. The results also show how, using the proposed approach, the computational tasks can be properly distributed between the fog network and a remote cloud server.

Motivation & Objective

• To address the challenge of dynamic fog network formation under uncertainty of neighboring fog node arrivals in IoT environments.
• To minimize the maximum computational latency in a hybrid fog-cloud architecture with no prior knowledge of future node availability or performance.
• To enable real-time, online selection of optimal fog nodes for task offloading without relying on offline knowledge of the system.
• To jointly optimize task distribution between local fog nodes and a remote cloud server under realistic queueing-based latency models.
• To develop a practical, scalable solution for fog computing that balances latency, cost, and dynamic network conditions.

Proposed method

• Formulates the fog network formation and task distribution problem as an online optimization problem under uncertainty.
• Adapts the online k-secretary framework to enable exploration of neighboring fog nodes before exploitation of the best candidates.
• Uses a queueing-theoretic model to accurately capture transmission and computational latency in both fog and cloud layers.
• Employs an online algorithm that observes node performance during an exploration phase and then selects optimal nodes for offloading during an exploitation phase.
• Integrates task distribution decisions between fog nodes and the cloud based on real-time latency and capacity metrics.
• Evaluates performance using competitive ratio analysis to compare online results against an offline optimal solution.

Experimental results

Research questions

• RQ1 How can a fog node dynamically form a low-latency network when neighboring nodes arrive unpredictably?
• RQ2 To what extent can an online algorithm achieve performance close to an offline optimal solution with full knowledge of future node arrivals?
• RQ3 How does the number of neighboring fog nodes affect task offloading distribution between fog and cloud layers?
• RQ4 What is the impact of node computational capability and distance to the cloud on task offloading decisions?
• RQ5 How effective is the proposed online algorithm in minimizing maximum computational latency under uncertainty?

Key findings

• The proposed online algorithm achieves a total cost within 2.7% of the offline optimal solution when six neighboring fog nodes are selected.
• The competitive ratio analysis shows that 16.2% of iterations achieve the exact optimal solution, and 50% achieve at least 94.2% of the optimal value.
• In the worst-case scenario across over 50,000 iterations, the competitive ratio was 0.59, indicating strong robustness.
• As the number of neighboring fog nodes increases from 6 to 7, total cost rises due to increased queue management overhead, though latency continues to decrease.
• When the distance between a fog node and the cloud increases from 200 m to 600 m, the percentage of tasks offloaded to the cloud decreases by 28.8% due to higher cloud latency.
• Improving fog node computational capability from μ=8 to μ=10 reduces cloud task offloading by up to 11.3%.