Solid-State Memory Data Randomization via Block Access Count Shifting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing solid-state memory technologies, such as NAND flash memory, face accelerated wear due to repeated programming of the same program states, leading to uneven distribution of data across memory cells, which can result in premature degradation.
Innovation Solution
A method and apparatus that randomize write data by shifting a multi-bit sequence value stored in a register based on the accumulated access count for each erasable block, combining it with the input data using a combinatorial function to ensure equal distribution of program states across memory cells, thereby reducing wear and enhancing data security.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data is repeatedly programmed to the same program states in NAND flash memory, then data storage functionality is maintained, but memory cell wear accelerates and device longevity decreases
Solution Approach 1:
The patent applies dynamics by making the data sequencing pattern variable and adaptive rather than fixed. The system dynamically adjusts the sequencing pattern based on program/erase cycle counts and wear characteristics of different memory blocks, allowing the data distribution pattern to evolve over time and adapt to changing wear conditions, thereby preventing repeated stress on the same program states
Solution Approach 2:
The patent changes parameters by modifying the data sequencing pattern parameters (such as shift amounts and permutation patterns) based on accumulated access counts and wear metrics. By varying these parameters over time, the system ensures that data is distributed across different program states in each programming operation, preventing accelerated wear while maintaining storage functionality
2Reliability
If data is uniformly distributed across memory cells, then wear is reduced, but data security against timing attacks may be compromised without proper randomization
Solution Approach 1:
The patent applies preliminary action by pre-generating randomization sequences and applying them to data before programming operations. This preliminary randomization ensures that even if timing information is available, the actual data patterns being programmed are unpredictable, thereby protecting against timing attacks while achieving uniform wear distribution
Solution Approach 2:
The patent introduces an intermediary randomization layer between the original data and the physical memory programming. This intermediary process uses pseudo-random sequences derived from wear metrics to transform the data pattern, creating a buffer that prevents direct correlation between timing information and underlying data patterns while maintaining uniform wear distribution
Data Source
AI summary
Method and apparatus for data storage. In some embodiments, a solid-state memory includes an array of non-volatile memory cells arranged into erasable blocks. A register stores a multi-bit sequence value. A controller randomizes input data to be written to a selected erasable block by combining the input data with the multi-bit sequence value shifted by a number of bit locations responsive to an accumulated access count for the selected erasable block.