ML-CaaS Framework for Anomaly Detection and Data Storage Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current AIOps solutions lack clarity in goals and milestones, inadequate infrastructure, and suboptimal utilization of machine learning across varied algorithms and methodologies, leading to inefficient anomaly detection and excessive historical data storage.
Innovation Solution
The ML-CaaS framework provides a dynamic anomaly detection system that proactively identifies anomalies using machine learning algorithms like isolation forest, reduces noise in alerts, and optimizes data storage by converting unstructured logs to structured datasets, thereby reducing storage requirements and improving real-time performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning algorithms are applied to anomaly detection in AIOps, then detection capability is improved, but storage requirements for historical data increase
Solution Approach 1:
The patent extracts only the essential features and patterns needed for anomaly detection from historical data, storing only relevant training data and model parameters rather than complete historical datasets. This selective extraction maintains detection capability while reducing storage burden.
Solution Approach 2:
The patent transforms raw historical data into structured training datasets with specific parameters and features optimized for machine learning algorithms. By changing the data representation from raw logs to structured training samples, the system achieves better detection with reduced storage requirements.
2Adaptability or versatility
If comprehensive AIOps capabilities are implemented, then system functionality is improved, but infrastructure complexity increases
Solution Approach 1:
The patent implements a unified ML-CaaS framework that provides multiple AIOps capabilities (anomaly detection, predictive analytics, root cause analysis) through a single standardized platform. This universal framework reduces infrastructure complexity by consolidating multiple functions into one system rather than requiring separate implementations for each capability.
Solution Approach 2:
The patent introduces an intermediary layer (the ML-CaaS framework) that sits between raw data sources and various AIOps applications. This intermediary provides standardized interfaces, data preprocessing, and model management, simplifying the infrastructure needed to support diverse AIOps capabilities.
3Measurement precision
If machine learning models are trained with extensive data, then model accuracy is improved, but training time increases
Solution Approach 1:
The patent extracts only the most relevant and informative features from training data, focusing computational resources on the most impactful data points. This selective feature extraction maintains model accuracy while reducing the computational burden and training time.
Solution Approach 2:
The patent applies partial action by using sufficient but not excessive training data - enough to achieve required accuracy thresholds while avoiding diminishing returns. The system dynamically determines optimal training set sizes based on performance requirements, balancing accuracy gains against training time costs.
Data Source
AI summary
A method comprises receiving data corresponding to execution of one or more applications, accessing at least one function from a codes as a service source, and training the at least one function based, at least in part, on one or more parameters, wherein the training is performed using a first portion of the data. In the method, a deployment version of the at least one function is generated based, at least in part, on the training, and the deployment version of the at least one function is applied to a second portion of the data to perform at least one service.