Network Interface Congestion Telemetry for Multi-Queue Bandwidth Allocation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing congestion control schemes in data centers fail to accurately manage bandwidth allocation across multiple traffic classes sharing an egress port, leading to inefficient transmission rate adjustments and potential over-allocation of bandwidth.

Innovation Solution

A network interface device calculates utilization for each traffic class based on queue depth, transmission rate, and egress bandwidth, providing congestion telemetry data to adjust transmission rates dynamically, allowing for differentiated Quality of Service and optimal bandwidth utilization across multiple queues.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If congestion control schemes propagate per-hop congestion states to senders, then congestion detection capability is improved, but bandwidth allocation accuracy deteriorates in multi-queue environments

Engineering Contradiction:
Improvecongestion detection capabilityVSAvoidbandwidth allocation accuracy
Core Design Contradiction:
Measurement precisionVSManufacturing precision

Solution Approach 1:

The patent segments the congestion measurement from the egress port level to the queue level. Instead of measuring congestion at the aggregate egress port, the system measures congestion separately for each queue that shares the egress port bandwidth. This segmentation allows accurate bandwidth allocation per traffic class while maintaining overall congestion detection capability.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies local quality by providing differentiated congestion information for different queues. Each queue receives congestion telemetry data specific to its own congestion state and bandwidth utilization, rather than a generic egress port congestion signal. This enables senders to adjust transmission rates based on the specific conditions of their traffic class.

Inventive Principle:
Principle #3Local quality

2Device complexity

If transmission rate is adjusted based on egress port congestion, then congestion control is simplified, but bandwidth allocation efficiency deteriorates across multiple traffic classes

Engineering Contradiction:
Improvecongestion control complexityVSAvoidbandwidth allocation efficiency
Core Design Contradiction:
Device complexityVSProductivity

Solution Approach 1:

The patent segments bandwidth allocation control to the queue level rather than applying a single egress port-wide control mechanism. Each queue can have its transmission rate adjusted independently based on its specific congestion state, improving overall bandwidth allocation efficiency while maintaining manageable complexity through automated telemetry collection.

Inventive Principle:
Principle #1Segmentation

3Manufacturing precision

If per-queue utilization is measured for congestion control, then bandwidth allocation precision is improved, but measurement complexity increases

Engineering Contradiction:
Improvebandwidth allocation precisionVSAvoidmeasurement complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent implements self-service by having the network switch automatically measure and report queue-level utilization metrics to senders. The switch monitors its own queue congestion states and generates telemetry data without requiring external measurement systems, thereby achieving precise bandwidth allocation while keeping the implementation complexity within the switch itself.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent establishes a feedback loop where the switch measures queue utilization and sends congestion telemetry data back to the senders. Senders use this feedback to adjust their transmission rates, creating a closed-loop control system that achieves precise bandwidth allocation through automated feedback rather than complex open-loop measurements.

Inventive Principle:
Principle #23Feedback

Data Source

PatentUS20230403233A1Congestion notification in a multi-queue environment
Publication Date: 2023.12.14 INTEL CORP
  • US20230403233A1 patent drawing
  • US20230403233A1 patent drawing
  • US20230403233A1 patent drawing

AI summary

Examples described herein relate to a network interface device. In some examples, the network interface device includes a host interface; a direct memory access (DMA) circuitry; a network interface; and circuitry. The circuitry can be configured to: based on received telemetry data from at least one switch: select a next hop network interface device from among multiple network interface devices based on received telemetry data. In some examples, the telemetry data is based on congestion information of a first queue associated with a first traffic class, the telemetry data is based on per-network interface device hop-level congestion states from at least one network interface device, the first queue shares bandwidth of an egress port with a second queue, the first traffic class is associated with packet traffic subject to congestion control based on utilization of the first queue, and the utilization of the first queue is based on a drain rate of the first queue and a transmit rate from the egress port.