Hybrid Network Assurance Model Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current network assurance systems face challenges in dynamically selecting appropriate machine learning models for monitoring networks, especially in hybrid architectures, where sensitive data cannot be sent to the cloud for analysis, and local resources are limited for training complex models.
Innovation Solution
A hybrid network assurance architecture where a local service reports configuration information to a cloud-based service, receives a classifier, and uses it to select a modeling strategy for installing a machine learning-based model for monitoring, allowing local assessment without sending confidential data to the cloud.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If machine learning models are trained locally using sensitive network data, then model accuracy and adaptability improve, but data security and privacy are compromised
Solution Approach 1:
The system segments the model training process into two distinct phases: cloud-based initial training using aggregated data from multiple networks, and local fine-tuning using minimal synthetic data. This segmentation allows the model to achieve high adaptability through cloud training while maintaining data security by avoiding transmission of sensitive local network data.
Solution Approach 2:
Synthetic data acts as an intermediary between the cloud training environment and local deployment. The synthetic data captures essential network characteristics and failure modes without containing real sensitive information, enabling the model to learn adaptive behaviors while preserving data security.
2Measurement precision
If complex machine learning models are deployed locally, then monitoring accuracy improves, but computational resource requirements exceed local capabilities
Solution Approach 1:
The system extracts and transfers only the essential model weights and architecture from the cloud-trained model to the local edge device. Complex training computations remain in the cloud, while the local device performs only lightweight inference and simple fine-tuning operations, achieving high monitoring accuracy without exceeding local computational resources.
Solution Approach 2:
The model is optimized for edge deployment by adjusting parameters such as model size, precision (e.g., using quantization), and computational complexity. These parameter changes enable the model to run efficiently on resource-constrained local devices while maintaining acceptable monitoring accuracy.
3Measurement precision
If cloud-based model training is used, then model performance improves, but latency in adapting to local network conditions increases
Solution Approach 1:
The cloud performs preliminary training of the model using aggregated data from multiple networks, pre-learning general network behaviors and failure patterns. This preliminary action enables the model to achieve good performance out-of-the-box, while local fine-tuning can quickly adapt to specific network conditions without requiring extensive retraining.
Solution Approach 2:
The system implements dynamic model updating where the model can be periodically retrained in the cloud with new data and deployed to local devices. This dynamic approach allows the model to adapt to changing network conditions over time, balancing cloud-based performance improvement with reduced adaptation latency through incremental updates.
Data Source
Figure 1A
Figure 1B
Figure 2
AI summary
In embodiments, a local service of a network reports configuration information regarding the network to a cloud-based network assurance service. The local service receives a classifier selected by the cloud-based network assurance service based on the configuration information regarding the network. The local service classifies, using the received classifier, telemetry data collected from the network, to select a modeling strategy for the network. The local service installs, based on the modeling strategy for the network, a machine learning-based model to the local service for monitoring the network.