AI Network Function Scheduling With Hierarchical Task Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing task scheduling for AI network function services lacks a common AI workflow and unified technical framework, leading to insufficient granularity in AI network function division and fragmented application scenarios, failing to meet users' personalized needs and efficiently schedule tasks.
Innovation Solution
A method for scheduling AI network function services involving a first network node selecting and sending tasks to multiple second network nodes based on their capabilities, using a deep reinforcement learning algorithm to optimize resource allocation and task division, and aggregating results for efficient and flexible AI service delivery.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If existing task scheduling methods are used for AI network function services, then the network can operate with current infrastructure, but the scheduling efficiency is insufficient and cannot meet personalized user needs
Solution Approach 1:
The patent segments the AI network function service into multiple granular task types (e.g., image recognition, speech recognition, natural language processing, recommendation algorithms). This segmentation enables differentiated scheduling strategies for different task types, improving both scheduling efficiency and the ability to meet personalized user needs by matching specific task requirements with appropriate network resources.
2Ease of operation
If AI network functions are divided into coarse-grained categories, then the system is easier to manage, but the granularity is insufficient leading to fragmented application scenarios
Solution Approach 1:
The patent implements multi-level task segmentation: first dividing AI network functions into major categories (image recognition, speech recognition, etc.), then further segmenting each category into specific task types with distinct characteristics. This hierarchical segmentation maintains system manageability at higher levels while achieving fine-grained control at lower levels, resolving the contradiction between ease of operation and task division granularity.
Solution Approach 2:
The patent applies different scheduling strategies and resource allocation methods to different task types based on their specific characteristics. For example, real-time tasks receive different treatment than batch processing tasks, and compute-intensive tasks are allocated differently from I/O-bound tasks. This localized optimization improves both manageability and granularity simultaneously.
3Adaptability or versatility
If a unified technical framework is implemented for AI workflow, then task scheduling can be standardized, but the complexity of the SDN and NFV architecture makes implementation difficult
Solution Approach 1:
The patent designs a universal task scheduling framework that can handle multiple AI task types (image recognition, speech recognition, NLP, recommendation algorithms) through a common architecture. The framework provides standardized interfaces and unified resource management while accommodating the diverse requirements of different task types, thus achieving workflow standardization without proportionally increasing architectural complexity.
Solution Approach 2:
The patent introduces an AI task management function as an intermediary layer between the SDN/NFV infrastructure and the diverse AI workloads. This intermediary translates various task requirements into standardized scheduling parameters, simplifying the interaction between the complex underlying architecture and the diverse upper-layer applications, thereby reducing the perceived complexity.
4Adaptability or versatility
If network resources are allocated to AI services, then personalized services can be delivered, but the resource allocation efficiency is insufficient
Solution Approach 1:
The patent dynamically adjusts resource allocation parameters based on task characteristics, user requirements, and network conditions. Different task types receive different resource quotas, priority levels, and allocation strategies. This parameter-based differentiation enables efficient resource allocation that simultaneously supports personalized service delivery and maintains high overall productivity by matching resource intensity to actual task needs.
Data Source
AI summary
A method for scheduling a task of an artificial intelligence (AI) network function service, performed by a first network node, includes: selecting at least two second network nodes from a plurality of second network nodes in response to a setup request for the AI network function service, wherein a node level of the plurality of second network nodes is less than a node level of the first network node; and sending a target task corresponding to the setup request for the AI network function service to the at least two second network nodes.