Temporal Relationship State Machine Observer for Multi-Core Fault Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional operation flow monitoring techniques in processing systems are inadequate for detecting hardware or software faults, particularly in multi-core systems, due to insufficient watchdog monitoring, difficulty in observing time constraints, and substantial overhead from in-line code monitoring, which limits their effectiveness in ensuring timely detection of hardware faults.
Innovation Solution
A state machine observer is implemented across multiple cores, using timed state machines to monitor execution conformance to temporal relationships, providing status indicators and independently detecting logical, control flow, and timing constraints, thereby enhancing fault detection and reducing dependent faults.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional in-line code monitoring is used, then fault detection capability is improved, but system overhead increases substantially
Solution Approach 1:
The patent extracts the monitoring function from the main execution flow by using separate watchdog processing units that independently monitor task execution. Instead of embedding monitoring code within the monitored tasks (in-line monitoring), the system uses external watchdogs that receive status information and generate alerts separately, thereby reducing the overhead on the monitored system while maintaining fault detection capability.
Solution Approach 2:
The patent introduces watchdog processing units as intermediary components between the monitored tasks and the fault detection system. These watchdogs act as mediators that receive execution status from tasks, evaluate temporal constraints, and generate fault alerts without requiring direct intervention in the task execution flow, thus reducing system overhead while improving reliability.
2Reliability
If one watchdog per processor core is used, then monitoring coverage is improved, but ability to monitor all tasks is insufficient
Solution Approach 1:
The patent implements a single watchdog processing unit that can monitor multiple tasks across different processor cores through the use of task identifiers and configurable monitoring parameters. The watchdog is designed to be universal, able to adapt its monitoring behavior based on the specific task being monitored, thereby providing both broad coverage and task-specific monitoring capability without requiring a dedicated watchdog per core.
Solution Approach 2:
The patent employs dynamic configuration of watchdog monitoring parameters, where the watchdog can adjust its monitoring behavior, thresholds, and target tasks based on system state and task requirements. This dynamic adaptability allows a single watchdog to effectively monitor multiple tasks with different temporal constraints and monitoring requirements, enhancing both coverage and versatility.
3Measurement precision
If individual time constraints between checkpoints are monitored, then timing accuracy is improved, but observation difficulty increases
Solution Approach 1:
The patent implements feedback mechanisms where watchdog processing units continuously receive execution status information from monitored tasks and compare actual timing against predefined temporal constraints. The system provides feedback alerts when constraints are violated, enabling precise timing monitoring while simplifying observation through automated constraint evaluation and alert generation rather than requiring direct observation of individual checkpoint timings.
Data Source
AI summary
A method includes receiving a first progress request from a first state machine associated with execution of a first thread on a processor. The method includes updating a current state of a temporal relationship state machine based on the current state, the first progress request, and a predetermined temporal relationship between progress of the first state machine to a first state machine state and progress to a second state. The predetermined temporal relationship may require the first state machine to progress to the first state machine state before the progress to the second state. The current state of the temporal relationship state machine may be one of a first temporal relationship state and a second temporal relationship state. The second state may be a second state machine state of the first state machine. The second state may be a second state machine state of a second state machine.