Abstract
Fog computing is an epitome to enable provisioning resources and services beyond the cloud, at the edge of the network, and nearer to end devices. Fog computing is not a counterfeit for cloud computing but a persuasive counterpart. However, the mobility of end users and/or fog nodes brings a major dilemma in the implementation of real-life scenarios. Therefore, we deliver the state-of-the-art research about mobility in fog computing. By identifying the mobility problems, requirements, and features of different proposals, we discover the open problems from subsisting studies and summarize the advantages of mobility for readers. This will help the researchers and developers avoid the current misapprehensions and capture the real-life scenarios, such as business, government, and education institutions. In spite of the extensive state-of-the-art research work, we also present the diverse mobility factors to investigate the correlation between dynamic nodes with fog nodes in order to accomplish better successful tasks. We conceive that the investigations of mobility factors in the fog computing will furnish the novel perspective on not only dynamic connections but also network dynamics. To revolutionize the follow-up research, we distinguish and foreground futurity directions concerning real-life scenarios of people and vehicles in a dynamic fog environment.
Introduction: Beyond the Centralized Cloud
While cloud computing offers massive power, its centralized nature leads to high latency—a dealbreaker for real-time applications like autonomous driving, augmented reality, and online gaming. Fog computing solves this by placing resources (nodes) closer to the user. However, fog is not a replacement for the cloud; it is a "persuasive counterpart" that extends the cloud's reach to the network edge.
The Fog Computing Paradigm
The architecture of fog computing is defined by a hierarchical structure that balances heavy processing with low-latency needs. This hierarchy is essential for maintaining service quality in complex urban environments.
As illustrated in the figure, the system is divided into three distinct planes: the Cloud Computing Plane for long-term storage and heavy computation, the Fog Computing Plane for localized processing, and the End-User Plane consisting of various smart devices, vehicles, and sensors.
Navigating the Mobility Dilemma
Early fog research often assumed a static environment. In reality, both users and fog nodes (e.g., a server inside a bus) are frequently in motion. This mobility creates a "dilemma": while it causes intermittent connectivity and task failures, it also provides "opportunistic contacts" that can be leveraged if predicted correctly.
Maintaining "link continuity" in these dynamic environments requires understanding how different entities move and interact across domains. The goal is to move from reactive connectivity to proactive, mobility-aware scheduling.
Modeling Dynamic Encounters
To address the gap between static models and urban reality, the authors categorize mobility into low-dynamic (Device-to-Device) and high-dynamic (Vehicular Fog) environments. They define three encounter types to model these interactions:
- Same Encounters: Synchronized proximity in both time and space.
- Partly Encounters: Shared location but at different time windows.
- Different Encounters: Random movement patterns compared via trajectory similarity.
Key Results and the "10-Second Flip"
Simulations involving 100 nodes (10 fog, 90 mobile) revealed critical performance insights regarding the Success Ratio—the percentage of tasks completed before a deadline:
- Proximity Advantage: In short-duration tasks, "Same Encounters" yielded success ratios 25% higher than "Partly Encounters."
- The Resilience of Randomness: Surprisingly, for tasks with a 10-second completion window, "Different Encounters" (random movement) outperformed synchronized meetings by 8%. This suggests that for delay-tolerant tasks, random mobility increases the statistical likelihood of eventually encountering any available node.
- Density Impact: Increasing the density of dynamic nodes significantly improves success ratios across all encounter types, as the probability of being within range of a fog node increases.
Future Directions and Open Problems
Despite progress, several hurdles remain for real-world implementation:
- Mobility Prediction: Utilizing machine learning to pre-allocate resources based on historical trajectories.
- Security & Privacy: Developing low-overhead authentication for transient nodes that may only stay in range for seconds.
- Incentive Models: Encouraging users to share resources (D2D) without depleting their own batteries.
- Fault Tolerance: Building systems capable of rapid recovery from "sporadic links" and intermittent interference.
Summary updated