Cloud Incident Reporting With Online and Fallback LLMs

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing cloud computing systems struggle with inefficiency, inaccuracy, and inflexibility in identifying, analyzing, and reporting service incidents and metric anomalies, often relying on obsolete methodologies that fail to accommodate various combinations of metric anomalies and providing unclear insights.

Innovation Solution

An anomaly mitigation proposal system utilizing a two-layer approach with online and fallback generative language models, where the online model generates text narratives for anomalous metrics, and the fallback model interfaces with root cause datastores to provide clear and concise incident reports when the online model is unavailable.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If rigid methodologies are used to identify and analyze service incidents, then the system structure is simple and easy to implement, but the system becomes obsolete and cannot accommodate various combinations of metric anomalies

Engineering Contradiction:

Improveability to accommodate various combinations of metric anomaliesVSAvoidsystem complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The system transitions from static rigid methodologies to dynamic adaptive methodologies that can evolve and adjust to different combinations of metric anomalies. The system continuously learns from new data patterns and updates its analysis capabilities, allowing it to adapt to emerging anomaly types without requiring complete system redesign.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system divides the complex task of anomaly analysis into multiple specialized components including machine learning models for pattern recognition, natural language processing for insight generation, and modular analysis engines that can be independently updated and scaled to handle different anomaly combinations.

Inventive Principle:

Principle #1Segmentation

2Reliability

If machine-learning models are employed to provide information about service incidents, then the system can handle various anomaly combinations, but the models are often unavailable and occasionally yield inaccurate solutions

Engineering Contradiction:

Improveaccuracy of incident analysisVSAvoidtime when models are unavailable

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The system implements fallback mechanisms and backup analysis methods that are prepared in advance for when machine learning models become unavailable. These cushioning measures ensure continuous operation with acceptable performance degradation rather than complete failure, maintaining system reliability during model unavailability.

Inventive Principle:

Principle #11Beforehand cushioning (Prior cushioning)

Solution Approach 2:

The system introduces intermediary components such as hybrid analysis methods that combine traditional rule-based approaches with machine learning outputs, and caching mechanisms that store previously analyzed patterns. These intermediaries bridge the gap during model unavailability and help validate or correct model outputs when available.

Inventive Principle:

Principle #24Intermediary (Mediator)

3Loss of information

If existing systems provide insights about service incidents, then some information is available, but the insights are suboptimal and unclear

Engineering Contradiction:

Improveclarity and quality of incident insightsVSAvoidefficiency of incident analysis

Core Design Contradiction:

Loss of informationVSProductivity

Solution Approach 1:

The system replaces traditional mechanical analysis methods with advanced computational approaches including natural language generation for insight formulation, semantic analysis for clarity improvement, and intelligent synthesis that combines multiple data sources to produce comprehensive and clear incident explanations.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system transforms the output parameters of incident analysis from simple metric reports to enriched insights that include natural language explanations, root cause hypotheses, impact assessments, and actionable recommendations. This parameter transformation significantly improves insight quality and clarity while maintaining analysis efficiency through automated processing.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS20260074948A1Anomalous metrics mitigation proposal system in a cloud computing system

Publication Date: 2026.03.12 MICROSOFT TECHNOLOGY LICENSING LLC

AI summary

The disclosure relates to utilizing an anomaly mitigation proposal system to determine root causes, summarize anomalous metrics, and report mitigation actions for service incidents in cloud computing systems. Based on receiving an incident report request, the anomaly mitigation proposal system utilizes a two-layer approach that implements large generative language models to generate incident reports that include clear and concise text narratives summarizing metric anomalies, root causes, and corresponding mitigation actions. For example, the anomaly mitigation proposal system initially utilizes an online generative language model to provide these incident reports and, when unavailable within a time threshold, a fallback model that references root cause datastores.