ML-Based EAS Ranking for 5G MEC Context Switchover Latency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing 3GPP 5G-MEC architecture faces challenges in maintaining low latency and high Quality of Service (QoS) during context switchover, particularly when multiple Edge Application Servers (EASs) are associated with a User Equipment (UE), as static notification sequences fail to meet Key Performance Indicators (KPIs) and result in low Quality of Experience (QoE) due to insufficient resource allocation.
Innovation Solution
A Machine Learning (ML) model is applied to rank Edge Application Servers (EASs) based on various parameters, determining an optimal notification order and resource allocation to ensure seamless context switching and improved QoS during data path changes, using parameters like QoS, computing load, network topology, and KPIs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a static notification sequence is used for context switchover, then the system complexity is reduced, but the latency requirement and QoS cannot be met when multiple EASs are associated with a UE
Solution Approach 1:
The notification sequence is changed from static to dynamic by using ML models to determine the optimal notification order based on real-time parameters such as QoS requirements, computing load, network topology, and KPIs. This allows the system to adapt the notification sequence dynamically to meet latency requirements while managing multiple EASs associated with a UE.
Solution Approach 2:
The system changes the notification sequence based on multiple parameters including QoS, computing load, network topology, and KPIs. By varying the notification order according to these parameters, the system optimizes latency performance without significantly increasing complexity, as the ML model handles the complex decision-making.
2Device complexity
If a static notification sequence is used, then the implementation is simpler, but the QoS and QoE are insufficient when the number of EASs increases
Solution Approach 1:
The notification sequence becomes dynamic and adaptive, determined by ML models that consider QoS requirements, computing load, network topology, and KPIs. This dynamic approach ensures reliable QoS performance even when the number of EASs increases, while the ML model abstraction keeps implementation complexity manageable.
Solution Approach 2:
The system uses feedback from multiple parameters (QoS, computing load, network topology, KPIs) to determine the optimal notification sequence. This feedback mechanism ensures that QoS requirements are met by continuously adapting the notification order based on current system state and performance metrics.
3Reliability
If multiple EASs are associated with a UE simultaneously, then service continuity is improved, but determining the notification order becomes more complex and latency increases
Solution Approach 1:
The ML model pre-processes and ranks EASs based on multiple parameters before the actual context switchover occurs. This preliminary ranking determines the optimal notification order in advance, allowing the system to handle multiple EASs efficiently without increasing notification processing time during the critical switchover moment.
Solution Approach 2:
The manual or rule-based mechanism for determining notification order is replaced with an ML-based system. This substitution enables the system to handle multiple EASs simultaneously while maintaining low latency, as the ML model can process and rank multiple EASs based on complex parameters much faster than traditional approaches.
4Reliability
If an ML model is applied to rank EASs, then QoS and QoE are improved, but the computational complexity and processing overhead increase
Solution Approach 1:
The ML model acts as an intermediary layer between the network entities and the context switchover process. It processes multiple parameters (QoS, computing load, network topology, KPIs) and outputs an optimized notification sequence, thereby improving QoS while managing computational complexity through a dedicated ML component rather than distributing complexity across all network entities.
Data Source
AI summary
The disclosure provides to a method and an apparatus for supervised learning approach for reducing latency during a context switchover in a 5G Multi-access Edge Computing (MEC). An example method for performing a context switchover in a wireless network includes identifying a plurality of first parameters associated with a user equipment (UE) and a plurality of second parameters associated with an edge network entity; receiving a data path change notification from a session management entity; determining a ranking for each of a plurality of edge application servers (EASs) based on the plurality of first parameters and the plurality of second parameters, in response to reception of the data path change notification; selecting at least one target EAS of the plurality of EASs based on the ranking for each of the plurality of EASs; and performing the context switchover to the at least one target EAS.