Distributed Spine-Leaf Network for Non-blocking IP Multicast
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional spine-leaf network architectures with controllers face high latency issues due to extra communication for node selection and bandwidth management, which is unsuitable for multimedia applications requiring low latency.
Innovation Solution
A distributed spine-leaf network architecture that uses a deterministic hash function for node selection and distributes bandwidth management across nodes, eliminating the need for a centralized controller by storing resource information on each switch, allowing for non-blocking IP multicast delivery of media data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a centralized controller is used for node selection and bandwidth management in a multi-spine network, then the network achieves centralized control and coordination, but the latency increases due to extra communication with the controller
Solution Approach 1:
The patent extracts the control function from a centralized controller and distributes it to individual switches. Each switch independently performs node selection and bandwidth management using locally stored resource information and deterministic hash functions, eliminating the need for communication with a central controller and thereby reducing latency.
Solution Approach 2:
Switches serve themselves by maintaining local copies of network resource information and autonomously making routing decisions. The deterministic hash function allows each switch to independently select spine nodes without external coordination, enabling self-service operation that eliminates controller communication overhead.
2Adaptability or versatility
If multiple spine nodes are used for scalability in a spine-leaf architecture, then the network capacity increases, but the complexity of node selection and bandwidth management increases
Solution Approach 1:
The patent uses copying by distributing identical network resource information to multiple spine nodes and leaf switches. Each switch receives and stores copies of bandwidth availability and node capability data, allowing them to independently make routing decisions without complex centralized coordination, thereby simplifying the management complexity while maintaining scalability.
Solution Approach 2:
The deterministic hash function transforms multiple possible routing options into a single selected path by changing the parameter selection process from complex multi-criteria optimization to a simple hash-based determination. This parameter transformation simplifies the node selection process while maintaining load distribution across multiple spine nodes.
Data Source
AI summary
In one illustrative example, an IP network media data router includes a spine and leaf switch architecture operative to provide IP multicast delivery of media data from source devices to receiver devices without the overhead communication with a controller. The architecture can include K spine switches, K sets of L leaf switches, M data links between each leaf switch, and a plurality of bidirectional data ports connected to each leaf switch for a guaranteed non-blocking IP multicast delivery of data. A deterministic hash function a used on both the first hop router and the last hop router to ensure the same spine node is selected for flow stitching. Accordingly, without the extra communication with a centralized controller, the right spine for establishing a multicast flow can be chosen using the deterministic hash function and the distributed resource information stored on each node.