Flash Storage Read-Threshold Tables for Low-Latency Data Retrieval
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current storage systems face performance degradation due to increased read latency caused by the need for error correction mechanisms when retrieving data from solid-state storage, as they often fail to retrieve error-free data efficiently.
Innovation Solution
Dynamic modification of read threshold values based on P/E cycle values, retention times, and page numbers to optimize data retrieval, allowing for more error-free data retrieval without the need for error correction mechanisms, thereby enhancing system performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If error correction mechanisms (ECC/RAID) are used to retrieve data from solid-state storage, then data reliability is improved, but read latency increases
Solution Approach 1:
The patent applies preliminary action by pre-calculating and storing optimal read threshold values in a read-threshold table before data retrieval operations. The system determines appropriate read thresholds based on program/erase cycle counts and page locations in advance, so that when data needs to be read, the correct threshold is already available, eliminating the need for error correction mechanisms and reducing read latency
Solution Approach 2:
The patent changes the read threshold parameter dynamically based on the program/erase cycle count and page location. By adjusting the read threshold voltage according to these parameters, the system optimizes data retrieval for different storage conditions, achieving reliable data reading without requiring error correction mechanisms
2Measurement precision
If dynamic read threshold modification is implemented based on P/E cycles and retention time, then data retrieval accuracy is improved, but system complexity increases
Solution Approach 1:
The system performs preliminary calculations to determine optimal read thresholds based on program/erase cycle counts and page locations, storing these values in a read-threshold table. This pre-computation approach simplifies the actual data retrieval process while maintaining high accuracy, as the complex threshold determination is done in advance rather than in real-time
Solution Approach 2:
The patent introduces a read-threshold table as an intermediary data structure that stores pre-calculated threshold values. This table acts as a mediator between the storage medium and the read operation, simplifying the interface and reducing the complexity of dynamic threshold adjustment by providing ready-to-use threshold values
Data Source
Figure 1A~1B
Figure 1C
Figure 2
AI summary
A method for reading data from persistent storage. The method includes : receiving (500) a client read request that includes a logical address for data from a client; determining (502) a physical address using the logical address where the physical address includes a page number for a physical page in the persistent storage; determining (504) a retention time for the data, determining (506) a program/erase (P/E) cycle value associated with the physical page; obtaining (508) at least one read threshold value using the P/E cycle value, the retention time, and the page number; issuing (510) a control module read request including the at least one read threshold value to a storage module that includes the physical page; and obtaining the data from the physical page using the at least one read threshold value.