Hybrid Application Probing for Root Cause Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Diagnosing the root causes of failures or performance issues in hybrid applications, which span multiple cloud environments and providers, is challenging due to the heterogeneous nature of these systems, leading to propagation of symptoms and inadequate diagnostic results from conventional tools.
Innovation Solution
An improved computing tool that utilizes a causal generative model and reinforcement learning to identify root causes, dynamically generates probes for data collection, and performs root cause identification across multiple cloud environments without pre-training, leveraging passive and active monitoring data to address monitoring data gaps.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional diagnostic tools are used on hybrid applications, then the diagnostic process is simple, but the diagnostic precision is insufficient due to symptom propagation across heterogeneous components
Solution Approach 1:
The patent segments the hybrid application into distinct functional modules and traces their interdependencies. By breaking down the monolithic diagnostic approach into module-level analysis, the system can identify which specific module causes symptoms to propagate across the heterogeneous architecture, thereby improving diagnostic precision without being overwhelmed by overall system complexity
Solution Approach 2:
The patent introduces a causal generative model as an intermediary that infers hidden causal relationships between application modules and infrastructure components. This mediator translates complex heterogeneous interactions into interpretable causal graphs, enabling precise root cause identification while abstracting away the underlying system complexity
2Reliability
If comprehensive monitoring is implemented across all hybrid application components, then the diagnostic capability is improved, but the resource expenditure increases significantly
Solution Approach 1:
The patent implements partial monitoring by focusing computational resources only on the causal paths relevant to the detected anomaly. Instead of continuously monitoring all components, the system dynamically identifies and monitors only the subset of modules and infrastructure components that are causally connected to the failure, maintaining high diagnostic capability while reducing resource expenditure
Solution Approach 2:
The patent performs preliminary causal inference using the generative model to predict which components are likely involved in the failure before deploying monitoring probes. This preliminary action filters out irrelevant components, so that subsequent comprehensive monitoring is applied only to the predicted causal subset, optimizing the balance between diagnostic capability and resource usage
3Measurement precision
If active probing is used to collect diagnostic data, then the measurement precision is improved, but the system performance degradation occurs due to additional overhead
Solution Approach 1:
The patent applies partial probing by selectively activating probes only on the causal paths identified by the generative model. Instead of deploying probes across the entire hybrid application, the system focuses probing efforts on the minimal subset of modules and infrastructure components that are causally connected to the anomaly, thereby achieving high measurement precision with minimal performance overhead
Solution Approach 2:
The patent dynamically adjusts the probing strategy based on the evolving understanding of the causal structure. As the generative model refines its inference about root causes, the system dynamically adds or removes probes to follow the discovered causal paths, ensuring that probing overhead is always minimized while maintaining measurement precision on the relevant components
Data Source
AI summary
Mechanisms are provided that detect an anomaly in performance of a hybrid application based on a specification of required performance and collected passive monitoring data, and that generate a causal generative model based on relationships between hybrid application components and computing system architecture components extracted from the passive monitoring data. Root cause identification (RCI) logic is executed on the causal generative model to identify a set of candidate root causes of the detected anomaly. One or more probes are identified for active monitoring data collection targeting the identified set of candidate root causes, which are then executed to collect probe data. Reinforcement learning is performed of the RCI logic to update the RCI logic based on the probe data. The set of candidate root causes is updated based on the reinforcement learning.