HeteroEdge: Latency-Aware Adaptive Protocol Parsing with Digital Twin Intelligence for Heterogeneous 5G IoT Edge...

Area of Science:

• Computer Science
• Networking
• Artificial Intelligence

Background:

• Rapid growth of heterogeneous IoT devices in 5G necessitates low-latency edge processing.
• Existing static parsing frameworks struggle with dynamic multi-protocol traffic, causing delays and Quality of Service (QoS) violations.
• Bursty workloads exacerbate processing challenges in current 5G Multi-access Edge Computing (MEC) environments.

Purpose of the Study:

• To introduce HeteroEdge, a novel latency-aware adaptive protocol parsing framework for 5G MEC.
• To address the limitations of static parsing in handling dynamic IoT traffic and reduce edge processing latency.
• To enhance the adaptability and efficiency of protocol parsing for latency-critical IoT applications.

Main Methods:

• Developed HeteroEdge, integrating a machine learning-based Heterogeneous Protocol Parsing Layer (HPPL) using XGBoost.
• Incorporated a Network Digital Twin (NDT) for real-time IoT endpoint state representation.
• Utilized a Real-Time Inference Engine (RTIE) for dynamic resource reallocation and a What-If Simulation (WIS) module for proactive strategy evaluation.

Main Results:

• HeteroEdge reduced median edge parsing latency by up to 44.7% compared to static MEC baselines.
• Achieved a macro-averaged protocol classification accuracy of 97.8% and sustained sub-7 ms latency at 18 Gbps.
• Reduced 95th percentile latency spikes by 39% and decreased SLA violation rates by a factor of 3.9 under bursty traffic.

Conclusions:

• HeteroEdge offers an effective and scalable solution for latency-critical IoT applications in 5G MEC.
• The adaptive framework significantly improves processing efficiency and QoS for dynamic and bursty IoT traffic.
• Demonstrated practical effectiveness on a physical 5G MEC testbed with diverse IoT endpoints and protocols.