Computational Storage NaN Replacement for Low-Latency Data Reads
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data storage systems require hosts to perform value replacement for missing values after data retrieval, leading to latency and computational overhead, especially when dealing with 'not a number (NaN) values.
Innovation Solution
The data storage device is configured to identify and store information about the locations of predetermined patterns, such as NaN, and perform on-the-fly replacement during read operations using a controller with integrated processors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the host performs value replacement after data retrieval, then data processing can be completed, but latency increases and computational overhead increases
Solution Approach 1:
The data storage device performs value replacement in advance during the read operation itself, before the host needs to process the data. The controller identifies NaN values and replaces them with default values (such as 0) as part of the read process, so that when the host receives the data, the replacement has already been completed. This eliminates the need for the host to perform separate replacement operations afterward, reducing both latency and computational overhead.
2Productivity
If the host performs value replacement after data retrieval, then data processing can be completed, but computational overhead increases
Solution Approach 1:
The data storage device performs the value replacement operation autonomously during the read process, without requiring the host to implement complex replacement logic. The controller automatically identifies NaN values and performs the replacement using default values defined in the system, effectively making the storage device serve the computational task of value replacement rather than simply storing and retrieving raw data.
3Reliability
If NaN values are stored as-is in the database, then data integrity is maintained, but additional processing is required by the host
Solution Approach 1:
The data storage device acts as an intermediary between data storage and data processing. Instead of storing raw NaN values and requiring the host to handle them, the storage device's controller performs the translation and replacement of NaN values with default values during read operations. This intermediary function absorbs the complexity of value replacement, allowing the host to receive cleaned data without needing to implement its own replacement logic.
Data Source
AI summary
A data storage device and method for data processing optimization for computational storage are disclosed. In one embodiment, a data storage device is provided comprising a memory and one or more processors. The one or more processors, individually or in combination, are configured to: receive data to be written in a data structure; analyze the data for a predetermined pattern; write the data in a plurality of storage locations in the memory; and write, in the memory, information about which storage locations store the predetermined pattern. Other embodiments are disclosed.


