Acceleration Resource Scheduling for NFV Latency Optimization

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional methods for selecting acceleration resources in NFV systems fail to ensure optimal latency and performance for services, as they only meet basic acceleration requirements and do not account for specific service needs.

Innovation Solution

An acceleration resource processing method that receives a service's acceleration resource request with attribute parameters and a service-specific scheduling policy, determining the appropriate acceleration resource based on these parameters to meet the service's requirements, including prioritizing local or remote virtualization and single-root I/O virtualization resources.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If acceleration hardware is selected based only on basic requirements (acceleration type and algorithm type), then the system can meet basic acceleration needs, but the service cannot achieve optimal latency and performance

Engineering Contradiction:
Improveservice latency and performanceVSAvoidresource selection complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent extends the resource selection criteria by adding scheduling policy parameters to the traditional acceleration type and algorithm type parameters. This allows the system to consider multiple dimensions (acceleration capability, latency requirements, performance standards) when selecting acceleration resources, thereby achieving optimal service latency and performance while managing complexity through structured parameter evaluation

Inventive Principle:
Principle #35Parameter changes

2Adaptability or versatility

If virtualization technology is used to pool and allocate acceleration resources dynamically, then network elasticity is improved, but it becomes difficult to ensure optimal acceleration effect for specific services

Engineering Contradiction:
Improvenetwork elasticityVSAvoidacceleration effect optimization
Core Design Contradiction:
Adaptability or versatilityVSManufacturing precision

Solution Approach 1:

The patent applies local quality by tailoring the acceleration resource selection to specific service requirements. Instead of uniform resource allocation, the system evaluates each service's unique scheduling policy and acceleration needs, selecting acceleration hardware that is locally optimized for that particular service's latency and performance requirements, thereby maintaining both elasticity and precision

Inventive Principle:
Principle #3Local quality

3Productivity

If acceleration resources are allocated without considering service-specific scheduling policies, then resource allocation is simplified, but service requirements such as latency and performance standards cannot be met

Engineering Contradiction:
Improveresource allocation efficiencyVSAvoidservice quality standards
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent implements preliminary action by pre-defining service acceleration resource scheduling policies that encode latency and performance requirements. When a service requests acceleration resources, the system already has the evaluation criteria ready, allowing for efficient resource matching that simultaneously achieves high allocation efficiency and meets service quality standards through pre-established selection rules

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3468151B1Acceleration resource processing method and apparatus
Publication Date: 2021.03.31 HUAWEI TECH CO LTD
  • EP3468151B1 patent drawingFigure 1
  • EP3468151B1 patent drawingFigure 2
  • EP3468151B1 patent drawingFigure 3

AI summary

The present invention provides an acceleration resource processing method and apparatus, and a network functions virtualization system. The method includes: receiving an acceleration resource request of a service, where the acceleration resource request includes an attribute parameter of an acceleration resource and a service acceleration resource scheduling policy, and the service acceleration resource scheduling policy is determined based on a service requirement of the service; and determining the acceleration resource of the service based on the attribute parameter of the acceleration resource and the service acceleration resource scheduling policy. According to this method, the acceleration resource is selected according to the service acceleration resource scheduling policy, so that a specific requirement such as latency sensitivity of a service can be met, thereby reducing a latency service and improving service performance.