Fuzzy Decision-Based Efficient Task Offloading Management Scheme in Multi-Tier MEC-Enabled Networks

Md. Delowar Hossain, Tangina Sultana, Md. Alamgir Hossain 0001, Luan N. T. Huynh, Junyoung Park, Eui-Nam HuhSensors 2021 · DOI

TL;DR

This paper introduces FTOM, a novel fuzzy logic-based system for efficiently managing task offloading in multi-tier MEC (Multi-access Edge Computing) networks. FTOM dynamically decides whether to offload tasks to local MEC servers, neighboring MEC servers, or the remote cloud, significantly improving task success rates and reducing latency compared to existing methods.

What Problem Does This Paper Address?

The rapid growth of data-intensive applications (like AR, IoT, and mobile gaming) in 5G networks places a huge burden on user devices and traditional cloud computing. While Mobile Edge Computing (MEC) brings computation closer to users, efficiently distributing tasks among multiple MEC servers and the cloud remains a challenge. Existing solutions often struggle with the dynamic nature of network conditions and user demands, leading to overloaded edge servers, increased task failure rates, and unacceptable latency. Specifically, previous approaches often focused on either vertical offloading (edge to cloud) or horizontal offloading (between neighboring edges), but rarely integrated both effectively, especially when considering the availability of underutilized neighboring MEC resources.

Methods

The core of the FTOM scheme is a fuzzy logic-based decision-making engine. This engine takes into account several key input variables:

  • Task Size ($\tau$): The computational requirement of the task.
  • Local MEC VM Utilization ($\iota$): The current load on the local MEC server.
  • Network Delay ($d$): The latency in the network.
  • Neighboring MEC VM Utilization ($\eta$): The load on nearby MEC servers.
  • WAN Bandwidth ($w$): The available bandwidth to the remote cloud.

These crisp inputs are first fuzzified into linguistic terms (e.g., "Small," "Medium," "Heavy," "Low," "High") using triangular membership functions. An inference engine then applies a set of fuzzy rules (IF-THEN statements) to determine the optimal offloading destination. Finally, a defuzzification step (using the Center of Gravity method) converts the fuzzy output back into a crisp decision, directing the task to either the local MEC server, a neighboring MEC server, or the remote cloud. The novelty lies in its ability to consider both local and neighboring MEC resources, as well as the remote cloud, in a dynamic and adaptive manner.

Datasets

The paper does not introduce new datasets. Instead, it uses simulation parameters based on existing literature [36,38,41] to model three types of applications: Augmented Reality (AR), Infotainment (I), and Health Monitoring (HM). These applications were chosen to represent a spectrum of requirements, from latency-sensitive (HM, AR) to delay-tolerant (I), and compute-intensive (AR).

Key Insights and Ideas

  • Leveraging Underutilized Resources: The most significant insight is the effective utilization of neighboring MEC servers. Instead of always defaulting to the remote cloud when a local server is busy, FTOM intelligently routes tasks to nearby MEC servers that might have spare capacity, thus improving efficiency and reducing latency.
  • Fuzzy Logic for Dynamic Environments: The paper demonstrates that fuzzy logic is a robust and low-complexity approach for handling the inherent uncertainty and dynamism of MEC environments. Its ability to work with linguistic variables and approximate reasoning makes it well-suited for making decisions under varying network conditions and server loads.
  • Holistic Decision-Making: FTOM's strength lies in its comprehensive consideration of multiple factors (task characteristics, local server load, neighboring server load, network conditions, and WAN bandwidth) to make an informed offloading decision, rather than relying on a single metric.
  • Balancing Latency and Resource Utilization: The scheme effectively balances the need for low latency (by prioritizing local/neighboring MECs for sensitive tasks) with efficient resource utilization (by offloading demanding or delay-tolerant tasks to the cloud when appropriate).

Key Results

The FTOM scheme demonstrated significant improvements across various performance metrics compared to benchmark schemes like LEO (Local Edge Offloading), TTEO (Two-Tier Edge Orchestration), FOLB (Fuzzy Orchestration-based Load Balancing), WOTO (Fuzzy Workload Orchestration-based Task Offloading), and FCTO (Fuzzy Edge-Orchestration based Collaborative Task Offloading):

  • Task Success Rate: FTOM improved the successfully executed task rate by approximately 68.5% compared to LEO, 32.4% vs. TTEO, 8.9% vs. FOLB, 3.2% vs. WOTO, and 38.6% vs. FCTO.
  • Task Completion Time: The scheme reduced task completion time by approximately 66.6% compared to LEO, 61.5% vs. TTEO, 47.9% vs. FOLB, 49.8% vs. WOTO, and 55% vs. FCTO.
  • Task Failure Rate: FTOM achieved a significantly lower average task failure rate, particularly under heavy load, with only 0.70% failure at 500 mobile devices compared to much higher rates for other schemes.
  • Congestion Handling: FTOM could handle up to 500 mobile devices without congestion, outperforming other schemes that experienced congestion at lower device counts.

Limitations and Future Work

  • Limitations: The paper acknowledges that the FTOM scheme's performance is evaluated through simulations. While the simulation parameters are based on existing literature, real-world deployment might introduce additional complexities. The paper also mentions that the fuzzy rules were empirically derived from existing studies, suggesting potential for further optimization.
  • Future Work: The authors propose exploring machine learning approaches for even more efficient task offloading in MEC-enabled networks. This could involve adaptive learning of fuzzy rules or entirely new ML-based offloading strategies.

Summary updated

BibTeX
@article{hossain2021fuzzy,
  author = {Md. Delowar Hossain and Tangina Sultana and Md. Alamgir Hossain and Md. Imtiaz Hossain and Luan N. T. Huynh and Junyoung Park and Eui-Nam Huh},
  title = {{Fuzzy Decision-Based Efficient Task Offloading Management Scheme in Multi-Tier MEC-Enabled Networks}},
  journal = {Sensors},
  year = {2021},
  doi = {10.3390/s21041484},
}
Figure