Smart Probing for Predictive Routing in LLNs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In bandwidth-constrained Low Power and Lossy Networks (LLNs), the high cost of probing for network information significantly impacts network performance, as it consumes precious bandwidth and does not scale well, especially when large numbers of probes are needed to retrieve performance metrics like delay measurements.
Innovation Solution
A machine learning-based 'smart' probing technique that dynamically selects nodes for probing based on relevance scores, tunes the probing rate according to algorithm convergence, accounts for critical traffic to minimize network impact, and allows node delegation of probing tasks to ancestors, thereby reducing unnecessary probing traffic.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional probing techniques are used to collect network information, then performance metrics can be obtained, but network bandwidth is consumed and performance deteriorates
Solution Approach 1:
The patent applies partial action by selectively probing only a subset of nodes rather than all nodes in the network. The system identifies and probes only those nodes that are most likely to provide valuable information for training the predictive model, thereby obtaining sufficient performance metrics while consuming minimal bandwidth.
Solution Approach 2:
The patent changes the parameter of probing frequency and target selection based on learned patterns. The system dynamically adjusts which nodes to probe and how often, transitioning from uniform probing to intelligent selective probing, thereby reducing bandwidth consumption while maintaining measurement accuracy.
2Loss of information
If large numbers of probes are sent to retrieve performance metrics, then comprehensive network information is obtained, but the probing cost increases significantly
Solution Approach 1:
The patent introduces an intermediary predictive model that mediates between the need for network information and the cost of probing. The model predicts which nodes are most informative and filters the probing targets, acting as an intermediary that reduces the number of probes needed while maintaining information completeness.
Solution Approach 2:
The patent performs preliminary action by using the predictive model to pre-identify valuable probing targets before actual probing occurs. This preliminary analysis allows the system to focus probes only on nodes that will provide the most information, avoiding wasteful probing of uninformative nodes.
3Reliability
If uniform probing of all nodes is performed, then all nodes are monitored, but unnecessary probing traffic is generated
Solution Approach 1:
The patent applies local quality by making different parts of the network subject to different probing strategies. Instead of uniform probing, the system applies targeted probing to specific nodes based on their predicted information value, creating a non-uniform probing distribution that optimizes both coverage and efficiency.
Solution Approach 2:
The patent introduces dynamics by making the probing strategy adaptive rather than static. The system continuously learns from probe results and adjusts its target selection, transitioning from a fixed uniform probing approach to a dynamic selective approach that responds to changing network conditions.
Data Source
AI summary
In one embodiment, network information associated with a plurality of nodes in a network is received at a device in a network. From the plurality of nodes, a node is selected based on a determination that the selected node is an outlier among the plurality of nodes according to the received network information. Then, a probe is sent to the selected node, and in response to the probe, a performance metric is received from the selected node at the device.