Distributed Load Aggregation for Information Processing Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In existing information processing systems, the load applied on determination of load distribution for aggregating state information across multiple information processing devices increases, leading to bottlenecks and potential disruptions in normal business operations, as the representative node becomes overwhelmed with CPU and network loads.
Innovation Solution
The system distributes aggregation processing across multiple nodes, allowing each node to determine whether to perform load distribution based on its own load and the loads of other nodes, without relying on a centralized database, thereby reducing the load on the representative node and preventing bottlenecks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If aggregation processing is centralized on one representative node, then data aggregation can be performed, but the load on the representative node increases excessively
Solution Approach 1:
The patent divides the centralized aggregation processing into multiple distributed aggregation processing units, each responsible for specific data aggregation tasks. Instead of one representative node handling all aggregation, multiple nodes perform aggregation independently, segmenting the overall workload and reducing the load on any single node while maintaining data aggregation capability.
2Ease of operation
If load distribution determination is performed centrally, then load balancing can be achieved, but the representative node becomes overwhelmed with CPU and network loads
Solution Approach 1:
Each information processing node independently determines its own load state and makes autonomous decisions about data aggregation and transmission. Nodes self-manage their aggregation processing based on local load conditions, eliminating the need for a centralized controller to make load distribution decisions, thereby reducing CPU and network load on any single representative node while maintaining load balancing.
3Loss of information
If all nodes transmit data to a single representative node, then centralized monitoring is achieved, but network bandwidth and processing capacity are bottlenecked
Solution Approach 1:
The patent segments the data transmission paths by creating multiple aggregation processing units distributed across different nodes. Each unit handles specific data aggregation and transmits results to other nodes or storage, creating parallel transmission paths instead of a single centralized path. This segmentation of transmission routes increases network bandwidth utilization and eliminates the bottleneck effect of centralized collection.
Data Source
AI summary
An information processing device included in a plurality of information processing devices of an information processing system, the information processing device including configured to acquire pieces of load information indicating loads on the plurality of information processing devices from each of the plurality of information processing devices, perform a first aggregation processing that aggregates the acquired pieces of load information, determine whether the first aggregation processing is to be distributed based on the acquired pieces of load information, when the information processing device determines that the first aggregation processing is to be distributed, transmit, to at least two information processing devices included in plurality of information processing devices, an execution request of second aggregation processing, the second aggregation processing being aggregation processing for each of a plurality of device groups obtained by dividing the plurality of information processing devices by the number of the at least two information processing devices.