Link Down Recovery via Sequence Numbering and Retransmission
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional computing systems lack mechanisms to survive link down events in point-to-point bus topologies, resulting in data loss and the need to restart applications or systems when a link between devices becomes impaired.
Innovation Solution
A method and system that maintain sequence numbering and flow control information during a link down event, allowing for the retransmission of 'in-flight' packets upon recovery, enabling the system to survive the event without restarting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional link down recovery mechanisms are used, then the system can recover from link failures, but all traffic on the link must be cleared and applications must be restarted, resulting in loss of transmitted data and system downtime
Solution Approach 1:
The patent applies preliminary action by pre-establishing alternative paths and maintaining connection state information before the link actually fails. When a link down event occurs, the system can immediately activate alternative routes and resume transmission without restarting applications, as the necessary state information was preserved in advance.
Solution Approach 2:
The patent implements discarding and recovering by selectively discarding only the failed link while recovering and maintaining the connection state information and alternative paths. This allows the system to recover from link failures without losing transmitted data or requiring application restarts, as the essential connection state is preserved and can be reused.
2Reliability
If link down events are treated as fatal errors with complete link restart, then the link can be reset, but system hang occurs and requires full system reboot
Solution Approach 1:
The patent applies segmentation by dividing the system into independent components that can fail and recover independently. When a link down event occurs, only the affected link is reset while the rest of the system continues operating. This prevents system-wide hangs and maintains productivity, as the failure is contained to a specific segment rather than propagating to the entire system.
Solution Approach 2:
The patent implements dynamics by enabling the link state to change adaptively based on failure conditions. The system dynamically adjusts the recovery process: detecting link failures, activating alternative paths, and resuming transmission without requiring a static full-system reboot. This dynamic response maintains system availability while ensuring reliable link recovery.
3Ease of operation
If bus arbitration schemes are implemented in shared topology, then device access to bus can be managed, but the arbitration schemes do not function properly as traffic on the bus increases
Solution Approach 1:
The patent applies the intermediary principle by introducing a shared switch as a mediator between devices and the bus. The switch manages traffic flow and arbitration centrally, preventing the arbitration schemes from failing under load. The intermediary switch handles the complexity of bus access control while allowing individual devices to communicate reliably, even when bus traffic increases significantly.
Solution Approach 2:
The patent implements mechanics substitution by replacing the mechanical bus arbitration scheme with an electronic switching system. Instead of relying on software-based arbitration that degrades under load, the system uses electronic switches that can dynamically route traffic with minimal overhead, maintaining reliability even when bus traffic increases.
Data Source
AI summary
A method for surviving a link down event in a two-way serial-connection link. The method includes sequentially numbering packets to be transmitted across the link, successfully transmitted packets being acknowledged and after the link down event. Only packets unacknowledged are retransmitted when a device coupled to the link attempts to survive the event, each unacknowledged packet being retransmitted in accordance with its number within the sequence.