PNIC Flow Selection for Elephant Flow Hardware Offload

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Physical network interface cards (PNICs) have limitations in processing flows using hardware acceleration techniques, and the network performance improvement is not maximized if flows that do not benefit significantly from hardware acceleration are offloaded.

Innovation Solution

Classify flows into elephant and mice flows, offloading packet processing for elephant flows to PNICs while processing mice flows in software, and use receive side scaling to map elephant flows to separate queues to avoid head-of-line blocking delays.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If hardware acceleration techniques are applied to process packets in the PNIC, then network performance is improved, but the PNIC has limitations around the number of flows that can be processed

Engineering Contradiction:

Improvenetwork performanceVSAvoidnumber of flows processed

Core Design Contradiction:

ProductivityVSAdaptability or versatility

Solution Approach 1:

The patent segments flows into two categories: elephant flows (high bandwidth, long-lived) and mice flows (low bandwidth, short-lived). This segmentation allows the system to apply hardware acceleration selectively to elephant flows while processing mice flows in software, thereby resolving the contradiction between improving network performance and managing the limited number of flows the PNIC can handle.

Inventive Principle:

Principle #1Segmentation

2Productivity

If flows are offloaded to the PNIC, then hardware acceleration is utilized, but network performance improvement is not maximized if flows do not benefit much from hardware acceleration

Engineering Contradiction:

Improvenetwork performance improvementVSAvoidinefficient hardware resource utilization

Core Design Contradiction:

ProductivityVSLoss of energy

Solution Approach 1:

The patent applies local quality by treating different flow types differently. Elephant flows receive hardware acceleration treatment (offloading to PNIC) while mice flows are processed in software. This selective approach ensures that hardware resources are concentrated on flows that benefit most from acceleration, maximizing network performance improvement while avoiding inefficient resource utilization.

Inventive Principle:

Principle #3Local quality

3Loss of time

If elephant flows are mapped to separate queues, then head-of-line blocking delays are avoided, but device complexity increases

Engineering Contradiction:

ImprovelatencyVSAvoidqueue management complexity

Core Design Contradiction:

Loss of timeVSDevice complexity

Solution Approach 1:

The patent segments the queue structure into separate queues for elephant flows and mice flows. This segmentation prevents head-of-line blocking where high-volume elephant flows would otherwise block low-latency mice flows. While this increases device complexity, it effectively reduces latency for time-sensitive traffic by ensuring that mice flows are not delayed by the processing of large elephant flows.

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS12445380B2Hardware acceleration techniques using flow selection

Publication Date: 2025.10.14 VMWARE INC

AI summary

In some embodiments, a method receives a packet for a flow associated with a workload. Based on an indicator for the flow, the method determines whether the flow corresponds to one of an elephant flow or a mice flow. Only when the flow is determined to correspond to an elephant flow, the method enables a hardware acceleration operation on the packet. The hardware acceleration operation may include hardware operation offload, receive side scaling, and workload migration.