Data Center Network Bottleneck Structures for Traffic-Aware Capacity

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing data center network designs, such as fat-trees and folded-Clos, are inefficient due to congestion-control algorithms not being adequately considered, leading to wasted resources and increased costs, and there is a lack of formal models for optimizing network configurations based on traffic patterns.

Innovation Solution

A mathematical model using Quantitative Theory of Bottleneck Structures (QTBS) is developed to optimize network designs by identifying optimal link capacities and switch configurations based on expected traffic patterns, ensuring minimal resource waste and efficient performance.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If full fat-tree topology is used, then network throughput is improved, but network cost increases due to excessive link capacities and switch resources

Engineering Contradiction:

Improvenetwork throughputVSAvoidnetwork resource waste

Core Design Contradiction:

ProductivityVSLoss of energy

Solution Approach 1:

The patent applies parameter changes by adjusting link capacities at different network levels based on traffic patterns. Instead of uniform full fat-tree link capacities, the invention optimizes link capacities to match actual traffic demands, thereby maintaining throughput while reducing resource waste and cost.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent introduces dynamic bottleneck identification and adaptation mechanisms that allow the network to adjust its effective capacity allocation based on real-time or expected traffic patterns. This dynamic approach enables the network to achieve optimal throughput without permanently provisioning excessive resources.

Inventive Principle:

Principle #15Dynamics

2Ease of manufacture

If homogeneous topology with identical switches is used, then ease of manufacture is improved, but adaptability to different traffic patterns deteriorates

Engineering Contradiction:

Improvenetwork deployment simplicityVSAvoidtraffic pattern optimization

Core Design Contradiction:

Ease of manufactureVSAdaptability or versatility

Solution Approach 1:

The patent applies local quality by allowing different link capacities and configurations at specific network levels and locations based on local traffic patterns. While switches remain homogeneous for ease of manufacture, the link capacities and bottleneck structures are customized locally to optimize for specific traffic flows and patterns.

Inventive Principle:

Principle #3Local quality

3Productivity

If link capacities are increased to handle peak traffic, then network throughput is improved, but network cost increases

Engineering Contradiction:

Improvenetwork throughputVSAvoidlink capacity resources

Core Design Contradiction:

ProductivityVSQuantity of substance

Solution Approach 1:

The patent applies partial action by provisioning link capacities that are sufficient for expected traffic patterns without over-provisioning for peak traffic that may not occur. The bottleneck structures are designed to handle the necessary traffic load while avoiding the cost of excessive capacity that would only be utilized during rare peak conditions.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS12413481B2Data center networks using bottleneck structures

Publication Date: 2025.09.09 QUALCOMM INC

AI summary

A network is designed based on its topology and the expected flow patterns in the network. The use of the latter can lead to efficient use of network resources and can reduce or even minimize waste. Non-interference properties of the expected flows can yield an improved or even optimal design.