Network Traffic Classification via ML-Based Slice Weighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current network slicing techniques rely on manual tagging and are inadequate for classifying traffic for newly released applications or those without pre-defined Quality-of-Service (QoS) requirements, as they cannot dynamically adjust network slices based on varying application operations.
Innovation Solution
Implementing a system that uses machine learning algorithms to classify network traffic into multiple slices based on application parameters, where each slice is associated with a weight, and adds these weights to packet headers for dynamic QoS modification, allowing for flexible selection of network links over a 5G cellular network.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If manual tagging based on static lists is used to classify applications to network slices, then classification is simple to implement, but it cannot handle newly released applications or applications without pre-defined QoS requirements
Solution Approach 1:
The system enables applications to self-classify by extracting parameters directly from application behavior and performance data, eliminating the need for manual tagging. The classification engine automatically analyzes application characteristics and assigns appropriate network slices based on observed performance metrics
Solution Approach 2:
The system performs preliminary classification by pre-defining multiple network slices with different QoS characteristics before application execution. When an application runs, the system proactively matches it to the most suitable slice based on real-time parameter analysis, preparing classification rules in advance
2Adaptability or versatility
If a single network slice is assigned to an application, then network management is simplified, but the application cannot dynamically adjust QoS based on varying operational requirements
Solution Approach 1:
The system segments network traffic into multiple classes by analyzing different parameters of the same application's packets. Instead of treating an application as a single entity, the system divides its traffic flow into multiple categories based on real-time performance requirements, allowing dynamic QoS adjustment for different packet types from the same application
Solution Approach 2:
The classification system transitions from static to dynamic operation by continuously monitoring application parameters and adjusting packet classification in real-time. The system adapts QoS requirements dynamically based on changing application state, network conditions, and performance metrics
3Reliability
If machine learning algorithms are used to classify packets dynamically, then QoS optimization is improved, but processing overhead and system complexity increase
Solution Approach 1:
The system applies machine learning selectively rather than to all packets uniformly. It uses ML algorithms for complex classification decisions while relying on simpler rule-based methods for routine packets, reducing overall processing overhead while maintaining QoS optimization for critical traffic
Data Source
AI summary
Systems and methods for classifying network traffic to use cellular network slices based on application parameters are described. In some embodiments, an Information Handling System (IHS) may include: a processor and a memory coupled to the processor, the memory having program instructions stored thereon that, upon execution, cause the IHS to: receive a plurality of packets originated by a single application; classify each of the plurality of packets into one of a plurality of network slices based upon network parameters of the application, where each of the network slices is associated with a weight, and for each given packet among the plurality of packets, add a weight to a header portion of the given packet, where the weight corresponds to the given packet's classification.