Fault Recognition in Distributed Real-Time Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Distributed real-time computer systems, particularly systems of systems (SoS), face challenges in efficiently recognizing and isolating faults to prevent their propagation, especially due to transient causes like cosmic radiation and software issues, which can impair data structures but are temporary, and existing methods lack effective parallelization of processing and fault recognition to improve response times.
Innovation Solution
The method involves a message distribution unit that associates a commitment time with messages, allowing parallel processing and fault recognition by relaying messages to monitor units immediately and delaying their transmission to processing units until the commitment time, enabling the rejection of faulty results before they affect the environment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If fault recognition is performed in the same processing unit, then system complexity is reduced, but fault tolerance deteriorates due to lack of independence
Solution Approach 1:
The system is divided into separate processing units and monitoring units. Each monitoring unit is assigned to monitor specific processing units, creating a segmented architecture where fault recognition is spatially separated from processing functions. This segmentation ensures that a fault in a processing unit does not compromise the fault recognition capability, as the monitoring unit operates independently.
2Reliability
If processing and fault recognition are parallelized in separate units, then fault tolerance improves, but response time deteriorates due to additional communication overhead
Solution Approach 1:
Monitoring units continuously monitor processing units in parallel with the execution of processing tasks. The fault recognition process is initiated simultaneously with processing operations rather than sequentially, allowing faults to be detected as they occur. This preliminary and parallel action eliminates the time penalty that would result from sequential fault checking after processing completes.
3Productivity
If messages are immediately relayed to processing units, then productivity is improved, but reliability deteriorates because faulty messages cannot be filtered
Solution Approach 1:
A message buffer acts as an intermediary between the message source and processing units. Messages are first stored in the buffer, where they can be validated and filtered by monitoring units before being relayed to processing units. This intermediary structure allows for reliable message filtering without blocking the overall message flow, as valid messages are quickly forwarded while faulty messages are identified and discarded.
4Reliability
If commitment time is enforced for all messages, then reliability improves by preventing faulty results, but productivity deteriorates due to processing delays
Solution Approach 1:
The commitment time mechanism is applied selectively rather than universally. Messages that are identified as valid and safe are relayed to processing units immediately without waiting for the full commitment time period. The commitment time constraint is enforced only for messages that require validation or show signs of potential faults. This partial application of the commitment time rule maintains reliability for critical messages while avoiding unnecessary delays for valid messages, thus preserving overall processing productivity.
Data Source
AI summary
A method for fault recognition in a distributed real-time computer system comprising fault containment units (FCUs), which has a global timebase, wherein the fault containment units communicate by means of messages via at least one message distribution unit, wherein a commitment time is associated with a message formed by a fault containment unit, and wherein a message distribution unit that receives a message relays the message to one or more fault containment units operating in parallel, and wherein a processing fault containment unit (VFCU) does not transmit or use any of its results that are influenced by one or more of the received messages to the environment of the processing fault containment unit or before the commitment times associated with the received messages.