Configurable Interconnect Data Processing System
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data processing systems, particularly those involving graphics processing units (GPUs), face challenges in efficiently managing multiple independent data processing functions, leading to complex and error-prone time-sharing approaches or increased costs due to separate units for each function.
Innovation Solution
A data processing system with a configurable interconnect and control circuitry that allocates sets of GPUs to virtual machines, allowing flexible and adaptable routing of data processing tasks, enabling transparent allocation and reallocation of GPUs without disrupting operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If a single data processing unit is time shared between multiple data processing functions, then the number of processing components is reduced, but the system becomes complex and error prone with poor performance
Solution Approach 1:
The patent segments the data processing unit into multiple independent execution contexts or virtual instances that can be allocated to different data processing functions. This allows a single physical GPU to be divided into multiple logical units that operate independently, resolving the contradiction by reducing the number of physical components while avoiding the complexity of traditional time-sharing through proper isolation and resource allocation mechanisms
2Quantity of substance
If a single data processing unit is time shared between multiple data processing functions, then the number of processing components is reduced, but execution independence is compromised
Solution Approach 1:
The patent implements segmentation by creating separate execution contexts or virtual instances within the data processing unit, each with its own memory space and execution state. This ensures that different data processing functions execute independently without interfering with each other, maintaining reliability while using fewer physical components
Solution Approach 2:
The patent introduces an intermediary resource allocation mechanism that manages the sharing of the data processing unit between multiple functions. This intermediary layer ensures independent execution by properly isolating resources and coordinating access, preventing conflicts while enabling multiple functions to share the same physical hardware
3Reliability
If a separate data processing unit is provided for each data processing function, then execution independence is ensured, but the number of processing components and cost increase
Solution Approach 1:
The patent merges multiple separate data processing units into a single shared unit that can be allocated to different functions. By implementing proper resource isolation and allocation mechanisms, the system achieves the execution independence of separate units while using fewer physical components, thus reducing cost and complexity
Solution Approach 2:
The patent implements a universal data processing unit that can serve multiple functions through flexible resource allocation. The single unit is designed to be multi-functional, capable of being allocated to different data processing tasks while maintaining execution independence through proper isolation mechanisms, eliminating the need for multiple dedicated units
Data Source
AI summary
A data processing system comprises one or more data processing units, a configurable interconnect and control circuitry. The control circuitry allocates one or more of the data processing units to a virtual machine and configures the configurable interconnect so as route one or more data processing tasks from the virtual machine to the one or more data processing units allocated for use by that virtual machine. This can provide a flexible and adaptable data processing system for carrying out the data processing tasks of a virtual machine, with the particular allocation of data processing units being substantially transparent to the virtual machine.