Centralized ML Predictor for Remote Network Management
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning (ML) models used for determining numerical similarity between text samples are computationally expensive, leading to resource-intensive processing that can slow down or fail other services in remote network management systems, such as web portals and server management.
Innovation Solution
Implementing a remote network management platform with separate computational instances for training, prediction, and end-user services, where the training instance generates ML models and the prediction instance applies them to determine numerical similarities, optimizing resource allocation and reducing latency in serving incident reports and knowledgebase articles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single server is used to provide both end-user services and ML model training/prediction, then device complexity is reduced, but the reliability and performance of end-user services deteriorate due to resource contention during ML operations
Solution Approach 1:
The system is divided into separate computational instances: end-user instances for providing web portals and managing networks, and dedicated training instances for ML model generation. This segmentation isolates resource-intensive ML operations from end-user services, preventing performance degradation while maintaining manageable system architecture.
Solution Approach 2:
A centralized machine learning predictor acts as an intermediary between end-user instances and the ML training infrastructure. End-user instances can submit prediction requests without directly running training operations, allowing ML model generation to occur in the background while services remain available to users.
2Speed
If ML models are trained locally on end-user servers, then prediction speed is improved, but the loss of computational resources increases due to duplicate training across multiple servers
Solution Approach 1:
Multiple end-user instances share a common trained ML model through the centralized machine learning predictor. Instead of each instance independently training models, the system merges training efforts into a single centralized operation, reducing redundant computational work while maintaining fast prediction capabilities across all instances.
Solution Approach 2:
The centralized machine learning predictor serves multiple end-user instances simultaneously, providing a universal prediction service. A single trained model can be applied to predict similarities across incident reports from multiple different end-user networks, eliminating the need for duplicate training while maintaining prediction speed.
3Productivity
If computational resources are dynamically allocated between training and prediction, then productivity is improved, but the device complexity increases due to resource management overhead
Solution Approach 1:
Computational resources are segmented into dedicated pools: training instances with resources optimized for model generation, and end-user instances with resources optimized for service delivery. This segmentation simplifies resource management by eliminating the need for dynamic allocation and load balancing, while maintaining high productivity through specialized resource pools.
Data Source
AI summary
A remote network management platform is provided that includes an end-user computational instance dedicated to a managed network, a training computational instance, and a prediction computational instance. The training instance is configured to receive a corpus of textual records from the end-user instance and to determine therefrom a machine learning (ML) model to determine the numerical similarity between input textual records and textual records in the corpus of textual records. The prediction instance is configured to receive the ML model and an additional textual record from the end-user instance, to use the ML model to determine respective numerical similarities between the additional textual record and the textual records in the corpus of textual records, and to transmit, based on the respective numerical similarities, representations of one or more of the textual records in the corpus of textual records to the end-user computational instance.