Network Probe Packet Pair Dispersion for Bottleneck Link Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing network probing methods, such as VPS and PGM, fail to accurately identify the bottleneck link in multi-link network routes, especially when layer-2 or layer-1 devices are involved, leading to underestimation of path capacity and increased complexity in pre-knowledge requirements.
Innovation Solution
A method using a pair of probe packets with different sizes (L1 ≤ L2) transmitted back-to-back to determine if the first link in the network route is the bottleneck by analyzing the dispersion of the packets across the network links, with the ratio of their sizes defining the indicator of time taken for the second packet to traverse the route.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If VPS probing is used to measure round-trip times for variable packet sizes, then capacity can be deduced from different RTTs, but the method fails to work correctly when the path includes multiple network links or switches, leading to underestimation of path capacity
Solution Approach 1:
The patent segments the network path into individual links by using back-to-back packet pairs and analyzing their dispersion at each hop. Instead of treating the entire path as a single aggregate, the method divides the measurement into link-specific components by tracking how packet pairs separate at each network element, enabling accurate capacity measurement of individual links within multi-link networks.
Solution Approach 2:
The patent introduces an intermediary measurement approach using packet pair dispersion as a mediator to infer link capacity. Rather than directly measuring capacity on multi-link paths, the method uses the dispersion behavior of packet pairs as an intermediary indicator that reveals bottleneck link characteristics without requiring direct access or control of intermediate network devices.
2Loss of time
If PGM probing uses two back-to-back probe packets of the same size, then transmission delay between packets is eliminated, but the method cannot identify which specific link is the bottleneck in a multi-link route
Solution Approach 1:
The patent applies local quality by making the two packets in each pair have different sizes (L1 and L2) rather than identical sizes. This local differentiation allows the measurement to capture specific characteristics of individual links along the path. The smaller packet can pass through links more quickly, creating a measurable dispersion pattern that reveals which specific link is the bottleneck, rather than just measuring aggregate path capacity.
3Measurement precision
If VPS probing is applied to individual links in a domestic network, then link-by-link capacity can be measured, but the pre-knowledge needed and method complexity increase considerably
Solution Approach 1:
The patent creates a universal probing method that can measure individual link capacity in domestic networks without requiring specialized pre-knowledge or complex configuration. The packet pair dispersion technique works across different network topologies and link types (layer-2 switches, layer-3 routers, access links) using a single standardized approach, eliminating the need for network-specific adaptations or extensive prior information about the network structure.
Data Source
AI summary
A method is described of probing a network route using a pairs of probe packets in which the first packets are typically smaller than the second. A first pair of probe packets with size ratio [L1/L2] and a second pair of probe packets with different size ratio [M1/M2] are transmitted onto a network route. An indicator of time taken for the second probe packet in the packet pairs to traverse the network route is derived and it is identified whether the values of the derived indicators vary or are substantially the same. The method, which can be used to probe home or domestic networks, indicates whether the first link in the route is or is not a bottleneck link, thereby providing further information about the route. A test apparatus, gateway device and computer to perform the method are also disclosed.