Trace Anomaly Grouping and Visualization for Root Cause Differentiation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional observability tools in software development and production environments have limited visibility into the internal state of executable code, making it difficult to diagnose malfunctions effectively, as they often rely on insufficiently detailed information collection and grouping errors by source location, which can mask the root cause of issues.
Innovation Solution
The trace anomaly grouping and visualization technique instruments executable code to capture detailed information about method invocations, data structures, and exceptions, logically grouping related traces to identify common causes of anomalies and differentiate between root causes, enabling comprehensive diagnosis and visualization of errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional observability tools collect information independently with insufficient detail, then the complexity of information collection is reduced, but the diagnostic capability and information fidelity are insufficient
Solution Approach 1:
The patent combines multiple independent information collection mechanisms (traces, logs, metrics, data structure snapshots) into a unified observability system. The trace sampling mechanism integrates with the data structure sampling mechanism to collect correlated information simultaneously, achieving high-fidelity diagnostic data without proportionally increasing system complexity.
Solution Approach 2:
The patent creates a universal sampling mechanism that can collect multiple types of information (traces, data structures, logs, metrics) through a common infrastructure. The same sampling infrastructure serves multiple diagnostic purposes, reducing overall system complexity while maintaining high information fidelity across different data types.
2Measurement precision
If errors are grouped by source location only, then the grouping process is simple, but the root cause differentiation capability is lost
Solution Approach 1:
The patent segments the error grouping process into multiple hierarchical levels: first grouping by source location, then further segmenting by data structure state, trace context, and exception details. This multi-level segmentation enables precise root cause differentiation while keeping each individual grouping operation relatively simple and manageable.
Solution Approach 2:
The patent adds multiple dimensions to the error grouping mechanism beyond just source location. Errors are now grouped across dimensions including data structure values, trace context, exception type, and temporal patterns. This dimensional expansion enables root cause differentiation without requiring a completely complex new system.
3Loss of information
If detailed trace information is collected for all exceptions, then the diagnostic information completeness is improved, but the data volume and processing overhead increase significantly
Solution Approach 1:
The patent applies partial sampling to trace collection, where not all traces are captured in full detail. Instead, a representative subset of traces is sampled based on criteria such as error frequency, criticality, and variability. This partial action approach maintains diagnostic information completeness for root cause analysis while significantly reducing the total data volume collected.
Solution Approach 2:
The patent extracts only the most relevant and discriminating information from trace data for collection and storage. Rather than capturing complete trace information for all exceptions, the system extracts key diagnostic elements such as critical data structure states, exception types, and contextual metadata that are most useful for root cause differentiation, reducing data volume while maintaining diagnostic effectiveness.
Data Source
AI summary
A trace anomaly grouping and visualization technique logically groups traces with anomalies to cases to enable software developers to monitor, diagnose and visualize the anomalies, as well as to solve the anomalies during application development and production. A client library of an investigative platform collects signals from traces (trace signal information). The technique organizes (groups) related trace signals of methods with anomalies (e.g., exceptions, performance abnormalities such as slowness) into datasets (denominated as “cases”) based on common cause for an anomaly and correlates the signals to identify a case. The collected information may be used to differentiate between root causes of the anomalies using a comparative visualization of traces displayed on a standard user interface of the investigative platform. As such, the technique facilitates an understanding of differences among traces of executable code that resulted in the failure and traces without failure by providing the ability to comparatively examine views of those traces displayed on the standard UI. Signals of two or more traces may be selected and displayed side-by-side for comparison. The traces may be selected from a general notion of a healthy trace and a failed trace.