Edge Workload Orchestration for Predictive 5G Function Scaling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems struggle to efficiently manage and scale network functions across edge sites in a telecommunications network, particularly in 5G networks, due to complex service chaining and unpredictable user demands, leading to potential service degradation or denial of service.
Innovation Solution
Implement a predictive model that forecasts network function usage spikes and dynamically adjusts instance allocation across edge sites and cloud resources, considering factors like connectivity, compute capacity, and bandwidth, using an orchestrator function and local edge managers to optimize resource distribution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If network function instances are scaled to meet user demands, then service availability and quality are improved, but resource consumption and operational complexity increase
Solution Approach 1:
The system performs preliminary actions by predicting future workload demands using machine learning models before the demands actually occur. This allows the orchestrator to proactively scale network function instances in advance, ensuring service availability is maintained while avoiding reactive complexity of last-minute scaling decisions
Solution Approach 2:
The system implements continuous feedback loops where workload metrics are monitored, predicted, and used to dynamically adjust instance allocation. The orchestrator receives feedback from edge sites about actual performance and resource usage, then refines scaling decisions to balance availability requirements with operational simplicity
2Reliability
If network function instances are scaled to meet user demands, then service quality is improved, but power consumption increases
Solution Approach 1:
The system applies dynamics by making instance allocation flexible and adaptive rather than static. The orchestrator continuously adjusts the number of running instances based on predicted workload patterns, ensuring that sufficient computing power is available to maintain service quality while minimizing power consumption during low-demand periods
Solution Approach 2:
The system changes operational parameters dynamically by adjusting instance allocation based on workload predictions. Machine learning models analyze historical and real-time data to predict when scaling is necessary, allowing the system to optimize the balance between service quality requirements and power consumption by only activating additional instances when genuinely needed
3Productivity
If workload distribution is optimized across edge sites, then resource efficiency is improved, but prediction accuracy requirements increase
Solution Approach 1:
The system segments the prediction problem by analyzing workload patterns at individual edge sites and then aggregating these insights for optimized distribution decisions. This segmentation allows the orchestrator to make accurate local predictions that collectively improve overall resource efficiency without requiring perfectly accurate global predictions
Solution Approach 2:
The system uses feedback mechanisms where actual workload outcomes are compared with predictions, and machine learning models are continuously refined based on this data. This feedback loop improves prediction accuracy over time while the system learns to make effective distribution decisions even with imperfect initial predictions, gradually reducing the gap between prediction requirements and actual performance
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Computing resources are managed in a computing environment comprising a computing service provider and an edge computing network. The edge computing network comprises computing and storage devices configured to extend computing resources of the computing service provider to remote users of the computing service provider. The edge computing network collects capacity and usage data for computing and network resources at the edge computing network. A predictive function is applied to the data to determine a predicted demand on the computing and network resources at a future time interval. Based on the predicted demand, a distribution of workloads is determined.