Dynamic Hardware Acceleration Allocation for Network Virtual Machines
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Cloud service providers face increased CPU utilization due to added features like encryption, leading to constrained resources, as they primarily rely on general CPU compute capabilities without utilizing purpose-specific hardware acceleration resources.
Innovation Solution
A system dynamically allocates purpose-specific acceleration units to virtual machines based on workload types, reducing CPU processing load by offloading tasks to specialized units like encryption, authentication, and compression acceleration units.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If general CPU compute capabilities are increased to handle increased demands, then processing capacity is improved, but resource constraints worsen and CPU utilization becomes constrained
Solution Approach 1:
The patent segments processing tasks by separating general CPU compute capabilities from purpose-specific hardware acceleration resources. Virtual machines are allocated to specific acceleration units based on workload types, dividing the processing system into specialized components that handle different task categories independently, thereby improving overall processing capacity without overloading the general CPU.
Solution Approach 2:
The patent creates a multi-functional processing architecture where the system can dynamically allocate workloads between general CPU resources and specialized hardware acceleration units. The virtual machine orchestration layer provides universal access to multiple types of processing resources, allowing the same infrastructure to handle diverse workload types including encryption, compression, authentication, and general computation.
2Adaptability or versatility
If features like encryption are added to applications, then functionality is improved, but CPU utilization increases and becomes constrained
Solution Approach 1:
The patent extracts specific computationally intensive functions from the general CPU and relocates them to dedicated hardware acceleration units. Encryption operations are offloaded to encryption acceleration units, authentication tasks to authentication acceleration units, and compression to compression acceleration units. This extraction reduces CPU utilization while maintaining enhanced functionality.
Solution Approach 2:
The patent introduces virtual machine orchestration as an intermediary layer between applications and hardware resources. This mediator automatically matches workload types with appropriate acceleration units, enabling applications to utilize specialized hardware resources for encryption and other intensive operations without directly increasing CPU utilization.
3Productivity
If purpose-specific hardware acceleration resources are utilized, then processing efficiency is improved, but resource allocation complexity increases
Solution Approach 1:
The patent implements self-service resource allocation where virtual machines automatically identify and allocate themselves to appropriate hardware acceleration units based on workload type. The system monitors CPU utilization and autonomously migrates workloads between general CPU and specialized acceleration resources without requiring manual intervention or complex external orchestration.
Solution Approach 2:
The patent creates a dynamic resource allocation system that continuously adapts to changing workload conditions. Virtual machines can be dynamically migrated between different processing resources based on real-time CPU utilization metrics and workload characteristics, allowing the system to optimize processing efficiency while managing allocation complexity through automated adaptation.
Data Source
AI summary
Technologies for dynamically allocating acceleration units of a network device include a network device configured to determine a present compute usage value associated with a workload of the virtual machine, determine whether to accelerate the virtual machine as a function of the present compute usage and a compute capability usage limit, and select, in response to a determination to accelerate the virtual machine, an acceleration unit from one or more acceleration units, as a function of a type of the workload. Additionally, the network device is configured to allocate the identified acceleration unit. Other embodiments are described and claimed.