Periphery Crossbars for Grid Routing Latency Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In interconnected processing engine grids, traffic distribution is often uneven, leading to congestion and latency issues due to local hotspots, making it difficult to achieve the shortest path with the lowest latency for data routing.
Innovation Solution
A system with an array of processing elements interconnected via a network on an integrated circuit, utilizing periphery crossbars connected to the edges of the array and storage components, which select lanes for routing data to ensure it travels in a straight line and minimizes hops, thereby reducing latency and congestion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If processing engines are interconnected via a mesh network in a grid configuration, then scalability with increases in the number of processing engines is improved, but traffic distribution becomes uneven leading to congestion and latency
Solution Approach 1:
The invention introduces periphery crossbars that segment the traffic flow into distinct lanes, separating high-traffic paths from lower-traffic paths. This segmentation allows traffic to be distributed more evenly across multiple routes, preventing congestion in any single path while maintaining the scalable grid architecture.
Solution Approach 2:
The invention adds a new dimension to the traditional mesh network by introducing periphery crossbars that create additional routing paths around the edges of the grid. This dimensional expansion provides alternative routes for traffic, enabling better load distribution and reduced latency without compromising scalability.
2Productivity
If traffic is routed through the network to take the shortest path, then routing efficiency is improved, but achieving the lowest latency becomes challenging due to hotspots
Solution Approach 1:
The invention applies local quality by configuring periphery crossbars to selectively route traffic based on local conditions. When a hotspot is detected in a particular region, the crossbars redirect traffic through alternative periphery lanes, ensuring that locally optimal routing decisions prevent congestion while maintaining overall routing efficiency.
Solution Approach 2:
The routing system becomes dynamic through the periphery crossbars, which can adaptively switch between different paths based on real-time traffic conditions. This dynamic routing capability allows the system to respond to changing traffic patterns and hotspots, maintaining low latency performance even when traffic demands fluctuate.
3Loss of time
If periphery crossbars are used to route traffic in straight lines with minimum hops, then latency is reduced, but device complexity increases
Solution Approach 1:
The periphery crossbars are designed with multi-functionality, serving both as routing switches and as traffic distribution mechanisms. By combining multiple functions into a single component, the invention reduces the need for additional dedicated structures, thereby limiting the increase in device complexity while achieving straight-line routing and minimum-hop paths.
Solution Approach 2:
The periphery crossbars act as intermediaries between the internal mesh network and external connections. This intermediary role allows them to manage traffic flow efficiently without requiring complex internal routing logic throughout the entire network, concentrating the routing intelligence in specific locations and simplifying the overall system architecture.
Data Source
AI summary
A system including an array of processing elements, a plurality of periphery crossbars and a plurality of storage components is described. The array of processing elements is interconnected in a grid via a network on an integrated circuit. The periphery crossbars are connected to a plurality of edges of the array of processing elements. The storage components are connected to the periphery crossbars.