Software-Distributed Global Load Routing in Hierarchical Switch Fabrics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing network architectures face challenges in scalability, versatility, and efficiency due to increasing network load and diverse traffic types from applications like high-performance computing and Internet of Things, with conventional congestion control mechanisms leading to slow responses and inefficient resource utilization.
Innovation Solution
A network architecture with flow channels and adaptive routing that uses software-generated global load information for dynamic flow setup and congestion control, allowing distributed management without a central controller, and incorporates Ethernet-compatible switches with high bandwidth and low latency, supporting diverse traffic types and reducing network hops.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional congestion control mechanisms are used, then network architecture simplicity is maintained, but network utilization and response speed deteriorate significantly
Solution Approach 1:
The network switches autonomously generate and distribute global load information without requiring a central controller. Each switch independently computes load metrics based on observed traffic patterns and shares this information with neighboring switches through standardized protocols, enabling self-organized congestion control that improves network utilization without centralized coordination complexity
Solution Approach 2:
The system implements feedback mechanisms where switches continuously monitor network conditions, generate global load information, and adjust routing decisions based on received load data from neighboring switches. This closed-loop feedback enables dynamic adaptation to changing network conditions, significantly improving throughput and reducing congestion without requiring complex centralized control
2Adaptability or versatility
If centralized controller is used for flow management, then coordination is simplified, but system scalability and single point of failure risk worsen
Solution Approach 1:
The patent extracts the centralized controller function and distributes it across multiple network switches. Each switch independently generates and processes global load information locally, eliminating the single point of failure and enabling scalable deployment. The control logic is extracted from a central authority and embedded within each switch, allowing parallel processing and fault tolerance
Solution Approach 2:
The control architecture is segmented into distributed modules where each switch independently manages its own flow control and routing decisions. Global load information is segmented and shared only with neighboring switches rather than requiring centralized processing, enabling modular scalability and reduced system complexity through distribution
3Productivity
If traditional routing algorithms are used, then routing simplicity is maintained, but application performance and throughput deteriorate under high network load
Solution Approach 1:
The routing algorithm dynamically adjusts paths based on real-time global load information exchanged between switches. Rather than using static routing tables, the system continuously monitors network conditions and re-routes traffic to optimize throughput. This dynamic adaptation enables the system to handle high network loads effectively by selecting optimal paths based on current congestion states
Solution Approach 2:
The system changes routing parameters such as next-hop selection and path metrics based on observed load conditions. Switches modify routing decisions by adjusting parameters like priority, bandwidth allocation, and path selection criteria in response to received global load information, enabling performance optimization under varying network conditions without requiring complex reconfiguration
4Adaptability or versatility
If network architecture is simplified, then implementation ease is improved, but handling diverse traffic types and applications deteriorates
Solution Approach 1:
The patent implements a universal global load information generation and distribution mechanism that handles diverse traffic types through a unified approach. The same basic protocol and data structures are used for different applications and traffic patterns, allowing the system to adapt to various workloads without requiring application-specific customization. This multi-functional approach maintains architectural simplicity while achieving high adaptability
Data Source
AI summary
Systems and methods are provided for performing routing in a switch network or fabric. Switches can be configured in a hierarchical topology having a plurality of groups, where switches in a group are connected to one another, and groups are connected to other groups. Routing can be performed by maintaining per-group group load information. A packet can be routed between at least two groups using the per-group group load information to effect a set of routing decisions. The set of routing decisions can be biased towards or away one or more paths.