Smart NIC Offloading Hypervisor Switching Latency

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

In software-defined infrastructure, the repeated routing of network data through two layers (ICM and hypervisor) leads to computational overhead, causing performance degradation for virtual machines during high network traffic, as the hypervisor shares finite computation capacity between network handling and VM processing.

Innovation Solution

Implementing Smart NICs (SNICs) that offload network switching and routing functions from the hypervisor to dedicated processing units, reducing the load on the ICM and allowing for pre-processing of network traffic before transmission, thereby enhancing performance and scalability.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If network switching and routing functions are handled by the hypervisor, then virtual machine management is achieved, but processing overhead and latency increase during high network traffic

Engineering Contradiction:
Improvenetwork processing throughputVSAvoidnetwork processing latency
Core Design Contradiction:
ProductivityVSLoss of time

Solution Approach 1:

The patent extracts the network switching and routing functions from the hypervisor and relocates them to dedicated Smart NICs. This separation allows the hypervisor to focus on VM management while the Smart NICs handle network processing, thereby reducing processing overhead and latency during high network traffic conditions.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent introduces Smart NICs as intermediary devices between the physical network and the virtual machines. These Smart NICs act as mediators that perform network switching and routing functions, reducing the burden on the hypervisor and improving overall network processing efficiency.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Productivity

If the hypervisor handles all network traffic processing, then comprehensive network control is maintained, but CPU availability for VM workloads decreases

Engineering Contradiction:
ImproveCPU availability for VM workloadsVSAvoidnetwork architecture complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent extracts network processing functions from the hypervisor and places them in dedicated Smart NICs. This extraction increases CPU availability for VM workloads by offloading network processing tasks, while the added complexity is managed through standardized interfaces and protocols.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The Smart NICs are designed to autonomously perform network switching and routing functions without requiring constant hypervisor intervention. This self-service capability reduces the computational burden on the hypervisor and improves CPU availability for virtual machine workloads.

Inventive Principle:
Principle #25Self-service

3Productivity

If basic NICs are used for network connectivity, then hardware simplicity is maintained, but network traffic pre-processing capability is lost

Engineering Contradiction:
Improvenetwork traffic pre-processing capabilityVSAvoidNIC functionality complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent implements Smart NICs that perform preliminary network traffic processing, including switching and routing decisions, before traffic is passed to the hypervisor or VMs. This pre-processing capability improves network efficiency by handling routine traffic management tasks in advance, reducing the workload on higher-level system components.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentUS11593140B2Smart network interface card for smart I/O
Publication Date: 2023.02.28 HEWLETT PACKARD ENTERPRISE DEV LP
  • US11593140B2 patent drawing
  • US11593140B2 patent drawing
  • US11593140B2 patent drawing

AI summary

A smart network interface card (SNIC) is provided. The SNIC may connect to an interconnect module (ICM) having at least two internal data paths. The SNIC and ICM determine a division of work between them. In general, NICs may be standard NICs, advanced NICs (ANICs), or smart NICs (SNICs). The ICM may perform a different amount of processing for network packets received from different devices based on the division of work previously identified. Some SNICs may preprocess network packets with respect to switching and routing processing to allow the ICM to bypass that functionality. Packets received from devices providing a division of work (e.g., SNICs) may receive reduced processing for functions offloaded to the SNIC. SNICs may utilize either a switching and routing group or a virtual bypass group such that data may bypass selected processing typically performed by the ICM.