Locating Packet Loss Sources in Distributed Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In large, distributed networks, identifying sources of packet loss, corruption, or latency has become increasingly challenging due to the complexity of software-defined networks (SDN) overlays and physical networks, despite each network component appearing healthy according to its own health model.
Innovation Solution
The solution involves collecting topology data for a packet-switched network, building a network topology of network nodes, tagging tracing packets, capturing packets from the network, identifying dropped tracing packets, and using the network topology to determine the last-visited network node, thereby generating a network performance report indicating the source of packet loss.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional network monitoring methods are used, then network troubleshooting can be performed, but the speed and precision of locating packet loss sources deteriorates due to increasing network complexity and SDN overlays
Solution Approach 1:
The patent introduces tracing packets as intermediary elements that carry identification information through the network. These packets act as mediators between the monitoring system and network components, enabling precise tracking of packet flow paths through SDN overlays and physical networks without being disrupted by their complexity
Solution Approach 2:
The system implements feedback by capturing tracing packets at monitoring points and comparing received packets against expected packets. This feedback mechanism identifies dropped or corrupted packets and traces them back to the last-visited network node, enabling precise localization of packet loss sources despite network complexity
2Reliability
If comprehensive packet capture is performed across the network, then packet loss can be detected, but the time and resources required for analysis increases
Solution Approach 1:
The patent extracts only the essential tracing packets from the network traffic stream for analysis. By filtering and capturing only packets with specific identification information rather than performing comprehensive packet capture, the system maintains reliable packet loss detection while significantly reducing analysis time and computational resources
Solution Approach 2:
The system changes the parameter of packet identification by embedding traceable identification information within tracing packets. This allows the monitoring system to reliably detect packet loss by checking for the presence or absence of specific identified packets, enabling rapid analysis without comprehensive capture of all network traffic
Data Source
AI summary
Solutions are disclosed that locate sources of packet loss in a distributed network. A network topology is constructed, and set of tracing packets is tagged. Packets are captured, including from host nodes (e.g., packet source and destination) and the tag is used as a filter to identify the tracing packets among the captured packets. The packet capture results are used to identify which (if any) of the tracing packets are dropped, and the network topology is used to identify each dropped packet's last-visited network node. This enables the generation of a network performance report indicating the location of the dropped packet(s) (if any). Some examples also include latency information in the network performance report.