Distributed GPU Service via MEC Resource Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current Multi-access Edge Compute (MEC) platforms face challenges in efficiently providing GPU services due to limited resource allocation and conservative sharing, leading to inefficient use of resources and inadequate support for graphics-intensive applications like virtual reality and augmented reality, which require high compute power and real-time processing.
Innovation Solution
A distributed GPU service that combines MEC GPU resources with end-device resources, allowing for parallel configuration of application data pipes and seamless fallback to local processing in case of connectivity issues, utilizing a provisioning framework to rapidly allocate virtual GPU resources across MEC clusters based on real-time utilization data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conservative resource allocation is used for shared GPUs in MEC platforms, then resource sharing is simplified and stability is improved, but resource utilization efficiency deteriorates and productivity decreases
Solution Approach 1:
The patent segments GPU resources into multiple virtual GPU instances that can be independently allocated and managed. Each virtual GPU can be assigned to different users or applications, enabling fine-grained resource sharing while maintaining stability through isolated resource management.
Solution Approach 2:
The patent implements dynamic resource allocation where GPU resources can be dynamically assigned, reassigned, or released based on real-time demand. This allows the system to optimize resource utilization efficiency while maintaining reliability through flexible resource management that adapts to changing conditions.
2Productivity
If more users are allocated to shared GPU resources, then resource utilization efficiency is improved, but system complexity increases and reliability deteriorates
Solution Approach 1:
The patent introduces virtualization layers and resource managers as intermediaries between physical GPU hardware and multiple users. These intermediaries abstract the complexity of shared resource management, enabling efficient multi-user allocation while simplifying the system architecture through standardized virtualization interfaces.
3Reliability
If GPU resources are allocated to meet peak demands, then service reliability is improved, but resource waste increases and productivity decreases
Solution Approach 1:
The patent implements resource reservation and pre-allocation mechanisms where GPU resources can be reserved in advance for specific users or applications. This allows the system to ensure service availability for critical users while avoiding resource waste by only allocating resources when actually needed, rather than maintaining peak capacity continuously.
Data Source
AI summary
Systems and methods provide a provisioning framework for a distributed graphics processing unit (GPU) service. A network device in a network receives, from an application, a service request for multi-access edge compute (MEC)-based virtual graphic processing unit (vGPU) services. The network device receives real-time utilization data from multiple MEC clusters in different MEC network locations and generates a utilization view of the multiple MEC clusters in the different MEC network locations. The network device selects, based on the real-time utilization view, one of the different MEC network locations to provide the vGPU services and instructs a of the multiple MEC clusters in the one of the different MEC network locations to perform container provisioning and service provisioning for the vGPU services.