Predictive Traffic Shaping for Network Reliability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing network optimization techniques rely on reactive approaches, leading to periods of reduced performance before corrective measures are taken, making it difficult to distinguish and optimize traffic flows for specific applications in enterprise networks with varying business and non-business critical traffics.
Innovation Solution
A predictive performance architecture that uses machine learning models to predict network resource availability, adjust traffic shaping strategies, and provide feedback for improved resource predictions, ensuring service level agreements (SLAs) are met dynamically.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If reactive techniques are used to enforce network performance criteria, then network performance can be monitored and adjusted, but periods of reduced performance occur before corrective measures are taken
Solution Approach 1:
The patent applies preliminary action by using machine learning models to predict future network resource availability and traffic patterns before performance degradation occurs. The system proactively adjusts traffic shaping parameters based on predictions, preventing performance issues rather than reacting to them after they occur. This eliminates the period of reduced performance inherent in reactive approaches.
2Productivity
If traditional traffic optimization methods are used, then network performance can be managed, but it is difficult to distinguish and optimize traffic flows for specific applications using the same protocols
Solution Approach 1:
The patent applies local quality by implementing application-specific traffic shaping parameters tailored to each application's requirements. Even when applications use the same protocols, the machine learning model identifies and applies differentiated shaping parameters based on application type, source, destination, and traffic characteristics. This enables precise optimization for each application while maintaining protocol compatibility.
Solution Approach 2:
The patent implements feedback mechanisms where the machine learning model continuously learns from actual network performance data and traffic patterns. The system monitors the effectiveness of applied traffic shaping parameters and adjusts future predictions and parameter selections based on this feedback, improving its ability to distinguish and optimize different traffic flows over time.
3Adaptability or versatility
If static traffic shaping parameters are used, then network configuration is simple, but the network cannot adapt to varying traffic demands and network conditions
Solution Approach 1:
The patent applies dynamics by implementing dynamic traffic shaping parameters that automatically adjust based on real-time network conditions and predicted resource availability. The machine learning model continuously updates shaping parameters in response to changing traffic patterns, network load, and resource availability, enabling the network to adapt to varying demands without manual reconfiguration.
Solution Approach 2:
The patent implements self-service through autonomous machine learning models that automatically predict network resource availability and determine optimal traffic shaping parameters without human intervention. The system self-adjusts to changing conditions, eliminates the need for manual parameter tuning, and continuously improves its performance through learned patterns from historical data.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A committed information rate (CIR) prediction (CIR-P) (502) is received by a traffic shaping module (249) from a network analysis module (NAM) (246). CIR-P (502) may be determined by NAM (246) in any number of different ways, depending on the learning machine techniques used by NAM (246). CIR-P (502) corresponds to a predicted average traffic rate supported by a network connection. A traffic shaping strategy is adjusted by the traffic shaping module (249) based on the CIR-P (502) wherein a rate at which data is communicated over the network connection is based on the traffic shaping policy. The effects of the adjusted traffic shaping strategy are monitored. Feedback is further provided to the machine learning model based on the monitored effects of the adjusted traffic shaping strategy. In an embodiment, traffic shaping module (249) may send a prediction request (504) to NAM (246) in order to explicitly request a CIR-P value from NAM (246). The request may be sent in response to detecting the presence of local queuing delays or in response to detecting an increase in dropped packets along the network connection. Illustrative embodiments that provide for predictive network control to be used in multicarrier wide area networks (WANs).