Dynamic Multi-Lane Link Degradation During Service Retrain
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing multiprocessor computer systems face challenges in maintaining uninterrupted service due to lane degradation in communication links, leading to potential loss of critical information and reduced bandwidth, as they require both nodes to independently identify lane degradation, causing temporary suspension of communications during link retrain.
Innovation Solution
A system and method for dynamically degrading links between processing devices, using a link retrain control module to determine undegraded lanes, set operational degrade modes, and dynamically select message packet types to ensure uninterrupted service, eliminating the need for simultaneous node input and utilizing spare lanes for critical traffic transmission.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If both nodes independently identify lane degradation before retrain, then link reliability is improved, but service interruption occurs during the identification and retrain process
Solution Approach 1:
The system performs preliminary identification of degraded lanes at one node before initiating retrain, rather than requiring both nodes to independently identify and agree on degradation. This preliminary action allows the retrain process to start immediately once degradation is detected, reducing service interruption time while maintaining link reliability through the subsequent retrain process
Solution Approach 2:
The patent introduces an intermediary mechanism where one node's identification of lane degradation serves as sufficient trigger for retrain, eliminating the need for both nodes to independently confirm degradation. This intermediary approach streamlines the decision-making process and reduces delays while preserving reliability through the retrain procedure
2Reliability
If spare lanes are reserved for degradation events, then bandwidth availability is improved, but the quantity of usable lanes is reduced
Solution Approach 1:
The system dynamically allocates lanes between service and spare functions based on real-time degradation status. Lanes that are not currently degraded are actively used for data transmission, while degraded lanes are automatically taken offline and replaced by spare lanes. This dynamic approach maximizes the utilization of available lanes while maintaining reliability, as spares are only activated when needed rather than being permanently reserved
Solution Approach 2:
The patent changes the operational state parameters of lanes dynamically - transitioning lanes between active service mode and spare mode based on their degradation status. This parameter change allows the same physical lanes to serve dual purposes: as active transmission channels when healthy, and as replacements when degraded, thereby increasing the effective number of usable lanes without sacrificing bandwidth availability
3Reliability
If link retrain is initiated to restore full functionality, then link performance is improved, but communication suspension occurs during retrain
Solution Approach 1:
The patent segments communication traffic into different priority levels, allowing critical communications to continue on undegraded lanes during the retrain process while non-critical traffic is suspended or redirected. This segmentation enables partial maintenance of communication throughput during retrain rather than complete suspension, while still restoring full link performance after retrain completes
Solution Approach 2:
The system maintains continuity of useful communication action during retrain by keeping undegraded lanes active and continuing to route traffic through them. Rather than suspending all communications during retrain, the system ensures that useful communication continues uninterrupted on healthy lanes, thereby maintaining productivity while still performing the necessary retrain to restore full link performance
Data Source
AI summary
A computer system, computer readable storage medium, and computer-implemented method for dynamically degrading a link between multiple processing devices to facilitate uninterrupted service between the multiple processing devices. The method includes determining a number of undegraded lanes within the link. The method also includes determining, subject to the number of undegraded lanes, a first operational degrade mode for the link. The method further includes initiating a retrain of the link to a second operational degrade mode. The method also includes dynamically, subject to the first degrade mode of operation, determining one or more message packet types to transmit through the link during the link retrain, thereby facilitating uninterrupted service between the multiple processing devices.


