Continuous Trace Logging for Scheduler-Level Process Metrics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In complex production environments, such as those for autonomous vehicles, identifying and correcting issues or bugs in the production environment can be challenging due to the difficulty in recreating conditions and the potential for time-sensitive corrections, especially when testing components are not present, which can lead to performance reductions and safety risks.
Innovation Solution
A production system with integrated data and process metric collection components that track and record metrics across multiple processing resources using a global clock, allowing for analysis and correction of issues without recreating them in a test environment, and transmitting logs to a remote debugging system for aggregation and analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Difficulty of detecting and measuring
If testing or data recording components are included in the production environment, then issue identification and debugging capability is improved, but system performance deteriorates due to consumption of processing resources
Solution Approach 1:
The system separates data collection functionality into independent modules (data collection component 104, process metric collection component 108) that can be selectively activated. These segmented components collect specific types of data (sensor data, process data, metric data) independently, allowing the system to maintain performance while enabling targeted debugging capabilities when needed.
Solution Approach 2:
The patent introduces intermediate data storage mechanisms (memory 106, remote storage 114) that buffer collected data without immediately processing it. This intermediary layer allows data to be captured and stored with minimal impact on real-time system performance, then analyzed later when performance constraints are less critical.
2Productivity
If testing components are removed from the production environment, then system performance is improved, but the ability to reproduce and debug issues deteriorates
Solution Approach 1:
The system performs preliminary data collection continuously in the production environment, capturing sensor data 120, process data 122, and metric data 124 before issues need to be debugged. By pre-collecting and storing this operational data with synchronized timestamps, the system ensures that debugging can proceed efficiently without needing to recreate production conditions.
Solution Approach 2:
The patent creates detailed copies of production environment data (process copies 118) that can be transmitted to remote systems for analysis. These copies include synchronized timestamp data that replicates the temporal relationships of the original production environment, enabling accurate debugging without requiring physical access to or recreation of the production system.
3Difficulty of detecting and measuring
If data collection and metric tracking are implemented across multiple processing resources, then issue analysis capability is improved, but system complexity increases
Solution Approach 1:
The patent implements universal data collection and synchronization mechanisms that work across multiple different processing resources (first processing resource 100, second processing resource 102). The global clock 110 provides a universal time reference that all resources use, and the synchronized timestamp generation works consistently across different resource types, simplifying the management of complexity through standardized interfaces.
Data Source
AI summary
Techniques for capturing and recording processor events and scheduler data in a production system on a per processing resource basis are discussed herein. In some examples, a process metric collection component may be associated with the scheduler and the processing resource such that the process metric collection component can capture real time data associated with the processes or threads both executed by the processing resource and waiting to execute on the processing resource. The captured data may be used by the system to monitor operations.