Virtualized CPU Resource Allocation for Cloud Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing cloud computing systems inefficiently allocate and track CPU accelerator resources, leading to suboptimal performance and inaccurate usage tracking for customers, as they often allocate resources without considering specific task requirements and fail to differentiate between users of accelerator resources and those who do not need them.
Innovation Solution
The system virtualizes CPU and on-chip enhanced CPU resources, using a virtualized-resource matching operation to allocate resources based on task requirements, ensuring that appropriate resources are allocated and usage is accurately tracked, thereby improving performance and efficiency by matching tasks with the necessary resources and preventing resource misuse.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If CPU accelerator resources are allocated without virtualization and task-specific matching, then resource allocation is simple and fast, but resource utilization efficiency deteriorates and task performance suffers
Solution Approach 1:
The patent segments CPU resources into virtualized CPU instances and virtualized accelerator instances, creating distinct allocable units. This segmentation enables fine-grained resource allocation where different virtualized components can be independently assigned to different tasks based on specific requirements, thereby improving resource utilization efficiency without requiring complete redesign of the allocation system.
Solution Approach 2:
The patent introduces a resource allocation system that acts as an intermediary between physical CPU resources and tasks. This intermediary virtualizes the resources and performs matching operations to allocate appropriate virtualized CPU instances and accelerator instances to tasks, simplifying the complexity by providing an automated mediation layer rather than direct management.
2Reliability
If general-purpose CPU resources are allocated to all tasks regardless of specific needs, then allocation process is straightforward, but task performance deteriorates due to mismatched resource capabilities
Solution Approach 1:
The patent segments CPU functionality into distinct virtualized instances including virtualized CPU instances and virtualized on-chip enhanced CPU instances. This segmentation allows the system to track and allocate specific functional capabilities to tasks, ensuring that tasks receive appropriate resource types (e.g., accelerator resources for compute-intensive tasks) thereby improving performance reliability.
Solution Approach 2:
The patent changes the parameter of resource allocation from generic CPU allocation to specific virtualized resource allocation with enabled/disabled functionality states. By virtualizing resources and tracking their specific capabilities and usage states, the system can match task requirements with appropriate resource parameters, improving reliability while managing complexity through structured parameter tracking.
3Measurement precision
If accelerator resources are shared without virtualization, then hardware cost is reduced, but resource allocation accuracy deteriorates and usage tracking becomes unreliable
Solution Approach 1:
The patent extracts the tracking and allocation logic from the physical hardware layer and places it in the virtualization layer. By virtualizing accelerator resources and creating separate virtualized instances, the system can precisely track usage at the software level without requiring additional physical tracking hardware, thereby achieving measurement precision while managing virtualization overhead through efficient software-based management.
4Adaptability or versatility
If physical hardware is used without virtualization to minimize overhead, then system simplicity is maintained, but resource allocation flexibility and adaptability deteriorate
Solution Approach 1:
The patent creates a universal virtualization layer that can allocate different types of resources (CPU instances, accelerator instances, enhanced CPU instances) to various tasks based on their specific requirements. This universal approach provides adaptability and versatility, allowing the same infrastructure to serve multiple different task types and workloads, while managing architecture complexity through a unified virtualization framework.
Data Source
AI summary
Embodiments of the invention provide a computer system that includes a central processing unit (CPU) associated with a host computer. The CPU includes CPU functionality and on-board enhanced CPU functionality. The CPU further includes a virtualized first instance of the CPU comprising an enabled virtualized first instance of the CPU functionality; and a non-enabled virtualized first instance of the on-chip enhanced CPU functionality. The CPU further includes a virtualized second instance of the CPU comprising an enabled virtualized second instance of the CPU functionality; and an enabled virtualized second instance of the on-chip enhanced CPU functionality.