Probabilistic Routing for Data Center Network Link Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing network routing technologies face challenges in optimizing cost and performance, particularly in data center networks, where they often result in inefficient link usage and looped paths that do not minimize network resources.
Innovation Solution
The implementation of a probabilistic routing system that uses probabilistic values to determine optimal paths for data packets across virtual local area networks (VLANs), incorporating a routing engine to generate probabilistic routing tables based on traffic measurements and network topology, and employing a backpressure algorithm to prioritize direct paths and minimize link usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional deterministic routing algorithms are used, then routing decisions are simple and deterministic, but link usage is inefficient and looped paths are created
Solution Approach 1:
The patent changes the routing parameter from deterministic to probabilistic by introducing randomization in the routing decision process. Each router selects a next hop based on probabilistic values rather than fixed deterministic rules, which breaks looped paths and improves link usage efficiency while maintaining manageable complexity through localized randomization
Solution Approach 2:
The routing system transitions from static deterministic routing to dynamic probabilistic routing where routing decisions can vary based on random selection. This dynamic approach allows the network to adapt to different traffic patterns and avoid congestion by randomly selecting among multiple possible paths, improving overall link utilization
2Adaptability or versatility
If probabilistic routing is implemented, then path diversity and link usage efficiency improve, but routing table complexity increases
Solution Approach 1:
The patent introduces probabilistic values as a new parameter in routing tables, allowing each entry to have an associated probability weight rather than a fixed deterministic choice. This enables path diversity by selecting different routes based on probabilistic values while keeping the routing table structure relatively simple and manageable
Solution Approach 2:
The routing decision process is segmented into independent probabilistic choices at each router. Each router independently selects its next hop based on local probabilistic values, distributing the complexity across multiple independent decisions rather than requiring a single complex centralized routing table
3Productivity
If deterministic routing is used, then routing implementation is simple, but network performance and cost optimization are limited
Solution Approach 1:
The patent changes the routing parameter from deterministic to probabilistic, enabling better network performance optimization. The probabilistic values allow for cost-aware routing decisions that can adapt to varying network conditions, improving performance while maintaining implementation simplicity through straightforward probability-based logic
Data Source
AI summary
A set of virtual local area networks (VLANs) are defined from a set of nodes that comprise a data center network. A set of possible paths are determined for routing data packets from a root node to one or more other nodes of the set of nodes. A set of probabilistic values are identified for determining a path in the set of possible paths that individual data packets are to take as between the root node and a destination node specified by the individual data packet. The data packets are routed to the destination node specified by that data packet, along a path that is determined by use of the set of probabilistic values. The path may be provided by one or more VLANs.