Virtual Function Accelerator Resource Scheduling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In virtualized environments, conventional schemes for utilizing accelerator resources often result in inefficient allocation and utilization, leading to decreased job execution efficiency due to the binding of accelerator resources to specific virtual machines, which limits flexibility and resource optimization.
Innovation Solution
A resource scheduling scheme is proposed that determines the job initiated by a virtual machine and allocates a first accelerator resource based on the job type, ensuring that each virtual function utilizes a single type of acceleration service, thereby optimizing resource allocation and improving execution efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If accelerator resources are bound to specific virtual machines, then resource allocation is simplified, but resource utilization efficiency deteriorates
Solution Approach 1:
The patent segments accelerator resources into multiple virtual functions (VFs), where each VF can be independently allocated to different virtual machines. This segmentation allows the system to maintain simple allocation management while improving utilization efficiency, as VFs can be dynamically assigned based on job requirements rather than binding entire accelerator devices to single VMs.
Solution Approach 2:
The patent implements dynamic resource allocation where accelerator resources are not statically bound to virtual machines but are dynamically assigned based on job types and requirements. The system can reallocate accelerator resources between different VMs depending on current workload demands, thereby improving overall resource utilization while maintaining manageable allocation complexity through automated scheduling.
2Productivity
If accelerator resources are dynamically allocated based on job types, then resource utilization efficiency is improved, but system complexity increases
Solution Approach 1:
The patent creates a universal scheduling framework that can handle multiple job types (compression, encryption, authentication, etc.) through a common resource allocation mechanism. Each virtual function is designed to be multi-functional, capable of handling different acceleration tasks, which reduces the need for separate specialized systems for each job type while maintaining high execution efficiency.
Solution Approach 2:
The patent introduces a scheduling system as an intermediary layer between virtual machines and accelerator resources. This mediator manages the complexity of dynamic allocation by abstracting the resource management logic, allowing job-type-based allocation without exposing complexity to the underlying VMs or hardware. The scheduler translates high-level job requirements into specific resource allocation decisions.
3Reliability
If virtual functions are configured to utilize accelerator resources, then service specialization is improved, but resource flexibility deteriorates
Solution Approach 1:
The patent segments accelerator resources into dedicated virtual functions, where each VF is configured for specific acceleration services (compression, encryption, etc.). This segmentation guarantees service performance by providing dedicated resources for each function while maintaining flexibility because the VFs themselves can be dynamically allocated to different VMs based on workload requirements, thus resolving the contradiction between specialization and flexibility.
Data Source
AI summary
A method, a device, and a computer program product for resource scheduling is disclosed. The method includes determining a job initiated by a virtual machine. The job requests to invoke at least one virtual function in a set of virtual functions associated with the virtual machine and each virtual function in the set of virtual functions is configured to utilize an accelerator resource to provide a single type of acceleration service. The method further includes determining, based on a job type of the job, a first accelerator resource allocated to the at least one virtual function. The accelerator resources required by the virtual functions invoked by the job may then be guaranteed, improving the execution efficiency of the job.