Dynamic Network-on-Chip Throttling via Detector-Mediator Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
As network-on-chip traffic increases, existing systems face performance degradation due to lack of effective traffic management, with agents having no direct knowledge of network state or capacity, leading to inefficient communication and potential overload.
Innovation Solution
Implementing a dynamic traffic throttling system where detector modules monitor queue occupancy and traffic conditions, sending signals to mediator modules to adjust traffic policies among agents, including thresholds for throttling policies such as 'light', 'heavy', and 'stop' to manage traffic generation based on outstanding transactions and service classes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traffic generation is increased to improve communication throughput, then productivity is improved, but performance degradation occurs due to network overload
Solution Approach 1:
The system implements feedback by having detector modules continuously monitor network conditions (queue occupancy, traffic load) and send signals to mediator modules, which then adjust traffic generation rates by agents. This closed-loop feedback mechanism allows the system to dynamically adapt traffic generation to current network capacity, preventing overload while maximizing throughput.
Solution Approach 2:
The patent applies dynamics by making the traffic generation rate adjustable and adaptive rather than fixed. The throttling policy dynamically changes the traffic generation rate based on real-time network conditions detected by detector modules, allowing the system to optimize throughput while avoiding performance degradation from excessive traffic.
2Reliability
If traffic throttling is implemented to prevent network overload, then network performance is maintained, but communication efficiency decreases due to reduced traffic generation
Solution Approach 1:
The throttling mechanism is dynamic rather than static, continuously adjusting the traffic generation rate based on real-time network conditions. When network performance is good, traffic generation is allowed to increase, improving communication efficiency. When performance degrades, throttling is applied to maintain reliability, creating an optimal balance between the two objectives.
Solution Approach 2:
The system changes the traffic generation rate parameter dynamically based on network conditions. The mediator module adjusts this parameter in response to detector module signals, allowing the system to optimize communication efficiency when possible while maintaining network performance through parameter adjustment when necessary.
3Device complexity
If agents generate traffic without knowledge of network state, then device complexity is reduced, but performance degradation occurs due to inability to manage traffic effectively
Solution Approach 1:
The patent introduces mediator modules as intermediaries between agents and the network. These mediators receive traffic generation requests from simple agents and adjust them based on network conditions, providing effective traffic management without requiring complex logic in the agents themselves. The detector modules also act as intermediaries, providing network state information to mediators.
Solution Approach 2:
The system segments traffic management functionality into separate modules: detector modules for monitoring, mediator modules for decision-making, and agents for traffic generation. This segmentation allows agents to remain simple while the overall system achieves effective traffic management through specialized modules working together.
4Reliability
If dynamic throttling policies are implemented to optimize performance, then network performance is improved, but device complexity increases due to additional detector and mediator modules
Solution Approach 1:
The system segments monitoring and control functions into separate detector and mediator modules, allowing each to be optimized for its specific function. This modular segmentation makes the overall complex system more manageable and maintainable, while the simplicity of individual modules offsets the added structural complexity.
Solution Approach 2:
The mediator module acts as an intermediary that simplifies the interface between simple agents and the complex network environment. By absorbing the complexity of network state interpretation and policy enforcement in the mediator layer, the system achieves effective performance optimization without requiring complex logic in every agent.
Data Source
AI summary
Dynamic network-on-chip traffic throttling, including: determining, by a detector module of a network-on-chip, that a predefined condition is met; sending, by the detector module, a signal to a mediator module of the network-on-chip; and sending, in response to the signal, by the mediator module, an indication to a plurality of agents to implement a traffic throttling policy.