Abstract
Recent advances in the areas of Internet of Things (IoT), Big Data, and Machine Learning have contributed to the rise of a growing number of complex applications. These applications will be data-intensive, delay-sensitive, and real-time as smart devices prevail more in our daily life. Ensuring Quality of Service (QoS) for delay-sensitive applications is a must, and fog computing is seen as one of the primary enablers for satisfying such tight QoS requirements, as it puts compute, storage, and networking resources closer to the user. In this paper, we first introduce FogPlan, a framework for QoS-aware Dynamic Fog Service Provisioning (QDFSP). QDFSP concerns the dynamic deployment of application services on fog nodes, or the release of application services that have previously been deployed on fog nodes, in order to meet low latency and QoS requirements of applications while minimizing cost. FogPlan framework is practical and operates with no assumptions and minimal information about IoT nodes. Next, we present a possible formulation (as an optimization problem) and two efficient greedy algorithms for addressing the QDFSP at one instance of time. Finally, the FogPlan framework is evaluated using a simulation based on real-world traffic traces.
Introduction
As Internet of Things (IoT) applications evolve toward autonomous drones, industrial robotics, and mobile augmented reality, traditional cloud computing architectures face significant challenges. These applications often demand latencies below 10ms, which the cloud frequently fails to provide due to unpredictable network hops and geographic distance.
Fog computing addresses this by distributing compute and storage resources across a "Fog Continuum" between the end-user and the cloud. However, fog nodes are resource-constrained compared to centralized data centers. Furthermore, IoT traffic is inherently "bursty"—a sudden event can flood local sensors with data while other areas remain idle. Static service placement leads to either expensive over-provisioning or performance-degrading under-provisioning.
The FogPlan Framework
To address the QoS-aware Dynamic Fog Service Provisioning (QDFSP) problem, we propose FogPlan. This framework dynamically manages the deployment and release of services on fog nodes to maintain low latency while minimizing operational costs.
At the heart of FogPlan is the Fog Service Controller (FSC). The FSC acts as a centralized intelligence that monitors network traffic and makes real-time provisioning decisions. It consists of several key components:
- Traffic Monitor: Aggregates incoming traffic rates from fog nodes.
- Provision Planner: Executes algorithms to decide where to deploy or release services.
- Service Database & SLA/QoS: Maintains information about available services and their specific performance requirements.
FogPlan is designed to be application-agnostic and lightweight. It does not require detailed knowledge of individual IoT device specifications or mobility patterns, relying instead on aggregated traffic metrics.
Modeling and Algorithms
The QDFSP problem is initially formulated as an Integer Nonlinear Programming (INLP) task. We utilize M/M/c queueing models to estimate waiting times on fog nodes and cloud servers, accounting for both processing delays and network propagation.
Because solving INLP is computationally prohibitive for large-scale, real-time networks, FogPlan employs two efficient greedy heuristics:
- Min-Viol (Minimum Violation): An aggressive QoS-focused approach that prioritizes deploying services on the busiest fog nodes to minimize latency violations.
- Min-Cost: A budget-conscious approach that only deploys a service to a fog node if the projected savings (from reduced cloud fees and avoided latency penalties) exceed the costs of local deployment and storage.
Experimental Evaluation
The framework was evaluated using a Java-based simulation environment driven by real-world traffic traces from the MAWI Working Group (April 2017). To test long-term scalability, we also utilized a traffic generator based on Discrete Time Markov Chains (DTMC) modeled after these traces.
Key results include:
- Near-Optimal Performance: In simulations, the Min-Viol algorithm achieved a delay violation rate of approximately 3%, performing remarkably close to the mathematically optimal INLP solution.
- Cost Efficiency: FogPlan significantly reduced total operational costs compared to an "All Cloud" approach by avoiding heavy financial penalties associated with missing latency targets.
- Scalability: The greedy algorithms scale quadratically. In tests with 10,000 fog nodes, the Min-Cost algorithm determined service placement in just 3.2 seconds, whereas traditional optimization methods would take hours.
Conclusion and Future Work
FogPlan demonstrates that dynamic, QoS-aware service provisioning in the fog is both practical and effective. By treating services as stateless entities (e.g., Docker containers), the framework can reconfigure the network in under 50ms.
Future work will address "impatient provisioning"—where the system reacts too quickly to short-lived traffic spikes—and explore the complexities of stateful applications requiring data migration. Additionally, incorporating machine learning for predictive traffic analysis could allow FogPlan to proactively move services before spikes occur.
Summary updated