Computational Storage Server Aggregating Client Inputs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current distributed computing systems are inefficient due to the need for each worker to lock and unlock shared files for matrix computations, leading to slow and non-scalable operations, especially in data-intensive applications like machine learning.
Innovation Solution
A computational storage server aggregates computation results from multiple clients, eliminating the need for workers to perform computations by locking shared files, and instead, performs arithmetic operations and softmax functions within the server, returning aggregated results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If each worker locks and unlocks shared files for matrix computations, then data consistency is maintained, but system performance and scalability deteriorate
Solution Approach 1:
The patent introduces a computational storage server as an intermediary between workers and shared files. The server performs computations on data stored in computational storage media, eliminating the need for workers to lock and unlock files. This mediator handles data consistency internally while allowing parallel access, thus maintaining reliability without sacrificing performance.
Solution Approach 2:
The patent replaces the mechanical file locking mechanism with a computational approach. Instead of using locks to ensure data consistency, the system uses computational storage servers that process data in memory and manage access control programmatically, enabling concurrent operations without traditional locking overhead.
2Manufacturing precision
If each worker performs computations by locking shared files, then computation accuracy is maintained, but operation speed and scalability worsen
Solution Approach 1:
The patent moves computations from the file system dimension to the memory dimension by implementing computational storage media. Data is processed in memory at the storage layer rather than being read from and written to files, adding a temporal dimension of parallel processing capability while maintaining computational accuracy through controlled access mechanisms.
Solution Approach 2:
The patent merges storage and computation functions into a unified computational storage server. This consolidation allows data to be processed directly where it is stored, eliminating the need to transfer data between storage and computation resources, thereby improving operation speed while maintaining accuracy through integrated error handling and consistency protocols.
3Reliability
If distributed computing uses traditional file locking, then data integrity is preserved, but system scalability deteriorates
Solution Approach 1:
The computational storage server provides universal access to multiple workers simultaneously without requiring individual locking operations. The server manages data integrity for multiple clients through a unified interface, allowing the system to scale by adding more workers without increasing locking complexity, thus improving scalability while preserving data integrity.
Data Source
AI summary
Computational storage techniques for distribute computing are disclosed. The computational storage server receives input from multiple clients, which is used by the server when executing one or more computation functions. The computational storage server can aggregate multiple client inputs before applying one or more computation functions. The computational storage server sets up: a first memory area for storing input received from multiple clients; a second memory area designated for storing the computation functions to be executed by the computational storage server using the input data received from the multiple clients; a client specific memory management area for storing metadata related to computations performed by the computational storage server for specific clients; and a persistent storage area for storing checkpoints associated with aggregating computations performed by the computation functions.