Service Application Learning for Packet Flow Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing service chaining solutions struggle with efficiently inserting or removing non-TCP-transparent services from data packet flows without interrupting end-to-end connections, especially when the traffic type is unknown or cached, leading to suboptimal service application and removal in service networks.
Innovation Solution
A method involving a target node classification entity that receives flow classification data, generates service classifiers, and updates a classification database to determine whether services should be applied to data packet flows, allowing for the adaptive steering of subsequent flows through or bypassing of service providing modules based on learned target node characteristics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traffic is sent through DPI first and then through non-TCP-transparent service, then service can process the traffic, but service cannot be removed from chain without interrupting connection
Solution Approach 1:
The patent performs preliminary classification of traffic flows through DPI before they reach the service chain decision point. By classifying flows in advance based on their characteristics (video, audio, data), the system can pre-determine the appropriate service chain configuration, enabling flexible adaptation while maintaining connection continuity for established flows.
Solution Approach 2:
The patent implements dynamic service chain configuration where the service chain is adapted based on traffic flow classification results. The system can dynamically insert or remove services from the chain depending on the classified traffic type, making the service chain flexible and adaptive rather than static, thus resolving the contradiction between connection reliability and chain flexibility.
2Reliability
If service is inserted into flow after connection establishment, then service can process traffic, but service cannot intercept TCP connection properly
Solution Approach 1:
The patent performs service chain configuration and service insertion decisions before TCP connection establishment is finalized. By classifying traffic and determining service requirements in advance, the service can be properly inserted into the connection path from the beginning, ensuring both connection stability and service processing efficiency without post-connection modifications.
3Adaptability or versatility
If all traffic is sent through service providing module, then service can process any traffic, but irrelevant traffic consumes service resources
Solution Approach 1:
The patent applies different service processing qualities to different traffic flows based on their classification. Video traffic receives video optimization services, audio traffic receives audio processing services, while data traffic bypasses these services. This local differentiation ensures that service resources are consumed only where needed, eliminating waste on irrelevant traffic while maintaining comprehensive processing capability for relevant traffic types.
Solution Approach 2:
The patent implements partial service application where only the necessary portion of service processing is applied to each traffic flow. Instead of applying all services to all traffic, the system applies only the specific service matching the traffic type (video service for video, audio service for audio), reducing unnecessary resource consumption while maintaining full service capability when needed.
4Measurement precision
If traffic classification is performed after connection establishment, then traffic type can be identified, but service cannot be inserted without interrupting connection
Solution Approach 1:
The patent performs traffic classification through DPI as a preliminary action during connection establishment, before the service chain needs to be configured. This early classification enables the system to identify traffic type and set up the appropriate service chain from the beginning, avoiding the need to interrupt connections later for service insertion while maintaining accurate traffic type identification.
Data Source
AI summary
The present invention relates to a method, by a target node classification entity (100), to control an application of a service providing module (50) to a data packet flow, wherein the data packet flow is transmitted to a target node (310, 320) through a service network where the service providing module (50) applies a service to the data packet flow. The method comprises the following steps: First, from a flow classification entity (40), flow classification data about the data packet flow is received, the flow classification data including at least a flow classifier classifying the data packet flow, a flow identifier uniquely identifying the data packet flow, and a target node address to which the data packet flow is addressed. Then, the target node is classified based on the flow classifier contained in the received flow classification data, wherein a first service classifier is generated for the target node based on the flow classifier. Then, it is checked whether a classification database (120) has to be updated based on the first classifier of the target node, the classification database storing, for different target nodes, whether at least one service has to be applied to data packet flows to the corresponding target node. Furthermore, if the information in the classification database has to be updated, a flow command is generated for a central flow controller (200) which controls the data packet flows through the service network, wherein the generated flow command contains the information that successive data packet flows in the service network transmitted to the target node address should be passed through the service network in accordance with the generated first service classifier of the target node, and the generated flow command is transmitted to the central flow controller (200).