Indirect Generalized Hypercube Network for Data Center Routing

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The upper layers of a fat tree network in data centers often become over-subscribed, unable to accommodate the maximum bandwidth demands of lower devices, leading to congestion and inefficient use of provisioned bandwidth.

Innovation Solution

Implementing an indirect generalized hypercube network (IGHN) within the data center, where each server has multiple network interfaces, allowing direct bi-directional links between servers in different racks, bypassing the aggregation layer and enabling alternative routing paths that reduce congestion and optimize bandwidth utilization.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If servers are connected through upper layers of fat tree network, then network connectivity is established, but bandwidth capacity is insufficient and congestion occurs

Engineering Contradiction:
Improvebandwidth capacityVSAvoidnetwork throughput
Core Design Contradiction:
Quantity of substanceVSProductivity

Solution Approach 1:

The patent segments the network routing paths by introducing direct inter-server links at the server level, separating the routing function from the traditional fat tree hierarchy. This creates multiple independent routing paths: through upper layers and through direct links, allowing traffic to be distributed and avoiding congestion at aggregation points.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces direct links between servers as intermediary pathways, bypassing the traditional upper layer routing. These direct links act as mediators that provide alternative routes for data traffic, reducing dependency on the over-subscribed upper layers of the fat tree network.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Productivity

If direct links between servers are added, then routing efficiency improves, but network complexity increases

Engineering Contradiction:
Improverouting efficiencyVSAvoidnetwork topology complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent implements direct links between servers selectively rather than universally. Not all servers need direct links, but strategic placement of direct links between frequently communicating servers provides sufficient routing efficiency improvement without requiring complete mesh connectivity, thus limiting the increase in network complexity.

Inventive Principle:
Principle #16Partial or excessive action

3Quantity of substance

If upper layers are over-subscribed, then bandwidth demand is high, but congestion occurs and bandwidth utilization becomes inefficient

Engineering Contradiction:
Improvebandwidth demandVSAvoidbandwidth utilization efficiency
Core Design Contradiction:
Quantity of substanceVSReliability

Solution Approach 1:

The patent extracts traffic from the over-subscribed upper layers by providing direct routing paths at the server level. This extraction reduces the load on upper layer links, allowing them to operate within their capacity while maintaining high bandwidth demand satisfaction through the new direct paths.

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentUS9929960B1Systems and methods for routing data through data centers using an indirect generalized hypercube network
Publication Date: 2018.03.27 GOOGLE LLC
  • US9929960B1 patent drawing
  • US9929960B1 patent drawing
  • US9929960B1 patent drawing

AI summary

Aspects and implementations of the present disclosure are directed to an indirect generalized hypercube network in a computer network facility. Servers in the computer network facility participate in both an over-subscribed fat tree network hierarchy culminating in a gateway connection to external networks and in an indirect hypercube network interconnecting a plurality of servers in the fat tree. The participant servers have multiple network interface ports, including at least one port for a link to an edge layer network device of the fat tree and at least one port for a link to a peer server in the indirect hypercube network. Servers are grouped by edge layer network device to form virtual switches in the indirect hypercube network and data packets are routed between servers using routes through the virtual switches. Routes leverage properties of the hypercube topology. Participant servers function as destination points and as virtual interfaces for the virtual switches.