5G MEC Traffic Offloading with SDN Latency Classes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing mobile networks face challenges in managing resource allocation for latency-sensitive applications, as centralized cloud execution introduces unacceptable latency, and multi-access edge clouds (MECs) may run out of resources, necessitating inefficient rerouting to a central cloud.
Innovation Solution
A resource manager orchestrates the redistribution of applications between MECs and a central cloud, using software-defined networking (SDN) to reroute latency-sensitive applications to available MECs or the central cloud based on resource thresholds and classifications, ensuring low latency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If applications are executed at centralized cloud services, then resource allocation is simplified, but latency increases to unacceptable levels
Solution Approach 1:
The patent segments the cloud infrastructure into multiple Multi-Access Edge Cloud (MEC) nodes distributed at different locations (e.g., base stations, regional centers). This segmentation allows applications to be executed closer to users, reducing latency while distributing resource management complexity across multiple nodes rather than centralizing it entirely.
Solution Approach 2:
The patent introduces a hierarchical dimension to cloud resource allocation, organizing MEC nodes into clusters that are grouped into regions. This multi-level hierarchy (nodes → clusters → regions) enables coordinated resource management across different spatial dimensions, allowing low-latency execution at edge nodes while maintaining centralized oversight for optimization.
2Loss of time
If MECs are deployed to reduce latency, then latency-sensitive applications improve, but MECs may run out of resources requiring inefficient rerouting to central cloud
Solution Approach 1:
The patent implements feedback mechanisms where MEC nodes and cluster heads continuously monitor resource usage levels and communicate this information to region heads. This feedback loop enables dynamic resource allocation decisions, allowing the system to detect when an MEC is approaching resource capacity and proactively redistribute work before outages occur, ensuring continuous availability.
Solution Approach 2:
The patent enables preliminary resource allocation and pre-positioning of applications at MEC nodes based on predicted resource usage patterns and current load conditions. By anticipating future resource needs and pre-allocating resources accordingly, the system prevents resource exhaustion scenarios and maintains continuous service availability without requiring emergency rerouting to central cloud.
3Ease of operation
If MECs are used for latency-sensitive applications, then user experience improves, but resource management complexity increases across multiple MECs
Solution Approach 1:
The patent merges resource management functions at the cluster level by introducing cluster heads that aggregate and coordinate resource allocation decisions across multiple MEC nodes within a cluster. This consolidation reduces the management overhead at individual nodes while maintaining distributed execution, effectively managing complexity through hierarchical coordination rather than centralized control of every individual node.
4Productivity
If applications are redistributed between MECs and central cloud, then resource utilization improves, but system complexity increases
Solution Approach 1:
The patent implements dynamic resource allocation where applications can be flexibly redistributed between MEC nodes and central cloud based on real-time conditions such as resource availability, latency requirements, and workload characteristics. This dynamic adaptability allows the system to optimize resource utilization by moving workloads between locations as conditions change, while the hierarchical management structure contains the complexity of these dynamic decisions.
Data Source
AI summary
A resource manager in a mobile communication network has a method to orchestrate execution of a first class of latency sensitive applications and a second class of latency sensitive applications where the second class of latency sensitive applications are less latency sensitive than the first class of latency sensitive applications. The method includes determining whether aggregated resource usage has exceeded or is predicted to exceed a threshold usage level for a first multi-access edge cloud (MEC), obtaining a list of the second class of latency sensitive applications executing at the first MEC, determining whether a second MEC in a cluster with the first MEC has resources to execute at least one application from the list of the second class of latency sensitive applications, and transferring a state of a selected application of the second class of latency sensitive applications to the second MEC.