Graph Neural Network Processing for Parallel Graph Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In systems where clients transmit data for arithmetic operations to servers, especially those using GPU clusters, resource management is inefficient due to the high computation costs and long processing times in tasks like neural networks, leading to suboptimal use of resources.
Innovation Solution
An information processing system that selects and processes multiple graphs simultaneously using a graph neural network model, optimizing resource allocation by determining the number of nodes and edges that can be processed concurrently within the server's capacity, allowing for parallel processing without interference and efficient resource utilization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple graphs are processed simultaneously to improve resource utilization, then productivity increases, but device complexity increases due to resource management overhead
Solution Approach 1:
The system dynamically changes resource allocation parameters by adjusting the number of nodes and edges processed simultaneously based on server capacity and task requirements. This allows optimization of processing throughput while managing resource complexity through parameter tuning rather than structural changes.
Solution Approach 2:
The patent implements dynamic resource allocation where the server adapts its processing capacity in real-time based on incoming task requirements and available resources. The system dynamically determines which graphs can be processed simultaneously by evaluating current server state, allowing flexible adjustment between productivity and complexity management.
2Speed
If more graphs are processed in parallel to reduce processing time, then speed increases, but loss of time increases due to resource contention and coordination overhead
Solution Approach 1:
The system processes a partial set of graphs in parallel rather than all available graphs, determining an optimal subset that can be processed simultaneously without causing resource contention. This partial action approach avoids the coordination overhead that would arise from managing too many parallel processes.
Solution Approach 2:
The patent ensures continuous processing by maintaining an optimal pipeline of graph processing tasks. When some graphs complete processing, new graphs are immediately assigned to freed resources, ensuring continuous useful action without idle time or excessive coordination overhead from starting and stopping parallel processes.
3Productivity
If server capacity is increased to handle more simultaneous graphs, then productivity improves, but loss of substance increases due to higher energy consumption
Solution Approach 1:
The patent segments the processing workload into multiple manageable graph tasks that can be distributed across available server resources. Rather than requiring a single large-capacity server, the system divides work into smaller units that can be processed in parallel by multiple smaller processing units, reducing total energy consumption while maintaining productivity.
Solution Approach 2:
The system makes server resources universal by allowing the same processing units to handle different graph processing tasks dynamically. This multi-functionality enables efficient resource utilization where the same hardware resources can be repurposed for different graphs based on current workload, avoiding the need for dedicated high-capacity resources for each task.
Data Source
AI summary
An information processing device includes one or more memories and one or more processors. The one or more processors are configured to receive information on a plurality of graphs from one or more second information processing devices; select a plurality of graphs which are simultaneously processable using a graph neural network model among the plurality of graphs; input information on the plurality of graphs which are simultaneously processable into the graph neural network model and simultaneously process the information on the plurality of graphs which are simultaneously processable to acquire a processing result for each of the plurality of graphs which are simultaneously processable; and transmit the processing result to the second information processing device which has transmitted the corresponding information on the graph.