Distributed Data Block Reduction for Stable Accumulation Order
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In distributed data processing systems, the stability of computation results is affected by differences in data collective logic among computing unit groups, leading to inconsistencies in data accumulation order and processing stability.
Innovation Solution
The method involves dividing data blocks into sub-blocks allocated to computing unit groups, ensuring a stable mapping relationship by using a data reduction operation on processing results within each group, particularly through ring all-reduce algorithms to maintain consistent accumulation order.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If data blocks are divided and allocated to multiple computing unit groups for parallel processing, then processing speed and computational power are improved, but data accumulation order becomes inconsistent and processing stability deteriorates
Solution Approach 1:
The patent segments data blocks into multiple sub-blocks and allocates them to different computing unit groups for parallel processing. Each computing unit group processes its assigned sub-blocks independently, enabling simultaneous computation across multiple units while maintaining overall processing stability through structured data distribution and collection protocols.
2Power
If computing units are organized into multiple computing unit groups, then computational capacity is improved, but differences in data collective logic cause accumulation order inconsistencies
Solution Approach 1:
The patent changes the parameter of data collective logic by implementing a standardized reduction operation protocol across all computing unit groups. This protocol ensures that regardless of which computing unit group processes which data sub-blocks, the accumulation and reduction operations follow a consistent logical sequence, thereby maintaining accumulation order consistency while utilizing multiple computing unit groups for enhanced computational capacity.
Data Source
AI summary
A data processing method is provided. The implementation is: for each of a plurality of data blocks included in data to be processed, dividing the data block into a plurality of first sub-blocks corresponding to a plurality of computing unit groups respectively; for each first sub-block of the plurality of first sub-blocks, dividing the first sub-block, based on a target computing unit group corresponding to the first sub-block, into a plurality of second sub-blocks corresponding to the plurality of computing units in the target computing unit group, respectively; determining, by processing each second sub-block utilizing the corresponding computing unit of the second sub-block, a plurality of first processing results output by the plurality of computing units respectively; and determining a processing result of the first sub-block by performing a data reduction operation on the plurality of first processing results utilizing the target computing unit group.