Distributed GPU Computation via Master Node Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional distributed computing systems are unable to handle the significant processing loads imposed by modern technologies, particularly autonomous driving systems, due to the lack of graphics processing unit (GPU) resources being available as a distributed resource.
Innovation Solution
A system and method for distributed graphics processing unit (GPU) computation using container-enabled systems, where a master node dynamically manages available computing resources and starts distributed processing containers with GPUs, and a load balancing server generates a unique URL for user tasks, allowing for efficient processing of GPU-intensive and CPU-intensive jobs across multiple computers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Power
If conventional distributed computing systems are used, then system simplicity is maintained, but the processing power for GPU-intensive tasks is insufficient
Solution Approach 1:
The system segments the distributed computing infrastructure into distinct node types (master nodes and worker nodes), where master nodes coordinate task distribution and worker nodes execute GPU-intensive computations. This segmentation allows the system to scale processing power by adding worker nodes without increasing the complexity of individual nodes, thereby resolving the contradiction between processing power and system complexity.
Solution Approach 2:
The patent introduces master nodes as intermediary components that mediate between task submission and GPU execution. These master nodes manage task queues, allocate work to worker nodes, and coordinate results collection, thereby enabling the system to handle complex GPU-intensive workloads without requiring each individual node to be overly complex.
2Productivity
If multiple task requests are assigned to a single computing system, then resource utilization improves, but processing speed deteriorates due to multitasking overhead
Solution Approach 1:
The system divides the computing workload across multiple independent worker nodes, each capable of processing tasks without interfering with others. This segmentation eliminates multitasking overhead between nodes while maintaining high resource utilization through coordinated task distribution managed by master nodes.
Solution Approach 2:
The patent combines multiple worker nodes into a coordinated distributed system managed by master nodes. This merging allows the system to achieve high resource utilization across all nodes while maintaining processing speed by having each node operate independently on assigned tasks without multitasking conflicts.
3Productivity
If distributed GPU resources are implemented, then processing capacity for image processing increases, but system complexity increases
Solution Approach 1:
The system segments GPU resources into discrete worker nodes, each with its own GPU capabilities. This segmentation allows processing capacity to scale by simply adding more worker nodes, while the complexity of individual nodes remains manageable. Master nodes coordinate these segmented resources without requiring complex integration between them.
Solution Approach 2:
The patent creates universal worker nodes that can handle various GPU-intensive tasks through standardized interfaces and task queues. This universality allows the system to increase processing capacity by adding identical multi-functional worker nodes rather than designing complex specialized systems for each task type.
Data Source
AI summary
A system and method for distributed graphics processing unit (GPU) computation are disclosed. A particular embodiment includes: receiving a user task service request from a user node; querying resource availability from a plurality of slave nodes having a plurality of graphics processing units (GPUs) thereon; assigning the user task service request to a plurality of available GPUs based on the resource availability and resource requirements of the user task service request, the assigning including starting a service on a GPU using a distributed processing container and creating a corresponding uniform resource locator (URL); and retaining a list of URLs corresponding to the resources assigned to the user task service request.