Network Controller Aggregates Packet Loss for Congestion Control

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional dual-mode congestion control algorithms fail to ensure fairness among dual-flows sharing a common network bottleneck due to inconsistent packet losses, leading to unfair bandwidth distribution and potential starvation.

Innovation Solution

A method where routers measure and aggregate packet losses across all flows at a bottleneck link, sharing this information with senders to enable coordinated switching between delay-based and loss-based congestion control modes, ensuring consistent packet loss data for all dual-flows.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional dual-mode congestion control algorithms are used, then each flow independently detects packet losses, but this leads to inconsistent packet loss information across dual-flows and unfair bandwidth distribution

Engineering Contradiction:
Improvepacket loss detection accuracyVSAvoidfairness among dual-flows
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The patent merges the packet loss detection function from individual flows into a centralized network controller. The controller aggregates packet loss information from multiple dual-flows sharing a common bottleneck, ensuring all flows receive consistent loss information for coordinated mode switching decisions, thereby resolving the fairness issue caused by inconsistent individual detections

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The network controller acts as an intermediary between the bottleneck router and the senders of dual-flows. It receives packet loss information from the router, processes and consolidates this data, then distributes consistent loss information to all relevant senders, enabling coordinated congestion control mode switching across all dual-flows

Inventive Principle:
Principle #24Intermediary (Mediator)

2Reliability

If centralized packet loss information is collected and distributed, then fairness among dual-flows is improved, but network controller complexity and information processing overhead increase

Engineering Contradiction:
Improvefairness among dual-flowsVSAvoidnetwork controller complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The system segments the congestion control functionality by separating packet loss detection (performed by the bottleneck router) from packet loss information distribution (performed by the network controller). This segmentation allows each component to focus on a specific task, reducing overall system complexity while maintaining fairness

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The bottleneck router automatically performs packet loss detection and reports this information to the network controller without requiring manual intervention. The controller then automatically distributes this information to relevant senders, enabling the system to self-manage the information gathering and distribution process

Inventive Principle:
Principle #25Self-service

Data Source

PatentUS11509595B2Network-based coordination of loss/delay mode for congestion control of latency-sensitive flows
Publication Date: 2022.11.22 CISCO TECHNOLOGY INC
  • US11509595B2 patent drawing
  • US11509595B2 patent drawing
  • US11509595B2 patent drawing

AI summary

A controller of a network, including routers to forward flows of packets originated at senders to receivers along distinct network paths each including multiple links, such that the flows merge at a common link that imposes a traffic bottleneck on the flows, receives from one or more of the routers router reports that each indicate an aggregate packet loss that represents an aggregate of packet losses experienced by each of the flows at the common link. The controller sends to the senders aggregate loss reports each including the aggregate packet loss so that the senders have common packet loss information for the common link on which to base decisions as to whether to switch from delay-based to loss-based congestion control modes when implementing dual-mode congestion control of the flows. In lieu of the controller, another example employs in-band router messages populated with packet losses by the routers the messages traverse.