Distributed Die ECC Layout for NAND Tile Failure Recovery
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing memory and storage systems face challenges in effectively mitigating media defects such as tile, plane, and die fails using error correction codes, particularly with distributed die error correction codes, where the XOR operation fails to reconstruct data due to lost codeword chunks, leading to excessive capacity loss from overdesigned XOR overhead.
Innovation Solution
Implementing inter-tile group (ITG)-XOR compatible die distributed codeword technology, where the ECC codeword is split and stored across multiple dies, with specific interleaving of codeword portions to ensure that the XOR operation can recover from tile fails by involving only one unknown in the XOR equations, thereby maintaining operational capacity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If distributed die ECC with XOR operation is used to correct data errors, then error correction capability is improved, but capacity loss increases due to overdesigned XOR overhead
Solution Approach 1:
The codeword is segmented into multiple portions and distributed across different dies. Specifically, the first codeword portion is stored in a first die and the second codeword portion is stored in a second die, allowing the XOR operation to be performed across distributed segments rather than requiring complete codewords in each die, thereby reducing the overhead required for error correction.
Solution Approach 2:
The error correction approach transitions from a single-die XOR operation to a multi-die distributed XOR operation. By distributing codeword portions across multiple dies and performing XOR operations across this new dimension (multiple dies), the system achieves better error correction capability with reduced overhead compared to traditional single-die approaches.
2Reliability
If complete codewords are stored in each die for XOR error correction, then error reconstruction is reliable, but storage efficiency decreases due to excessive XOR overhead
Solution Approach 1:
The complete codeword is segmented into first and second portions that are distributed across different dies. This segmentation allows each die to store only a portion of the codeword rather than complete codewords, reducing the redundancy overhead while maintaining the ability to reconstruct data through XOR operations across the distributed portions.
Solution Approach 2:
The system merges the distributed codeword portions from multiple dies through XOR operations to reconstruct the complete codeword. By combining the first codeword portion from one die with the second codeword portion from another die, the system achieves reliable error correction without requiring each die to store complete redundant codewords.
3Reliability
If XOR operation involves multiple unknowns from lost codeword chunks, then error correction coverage is expanded, but solution uniqueness is lost leading to failed reconstruction
Solution Approach 1:
The codeword is segmented and distributed such that the XOR operation involves only a single unknown variable. By storing the first codeword portion in one die and the second codeword portion in another die, and designing the XOR operation accordingly, the system ensures that when errors occur, the reconstruction equation has exactly one unknown, making it solvable and maintaining ease of operation.
Data Source
AI summary
An embodiment of a semiconductor apparatus may include technology to store a first portion of a code for a tile in a first die of the two or more nonvolatile memory die, store a second portion of the code for the tile in a second die of the two or more nonvolatile memory die, and perform an exclusive-or operation to correct a data error in the tile based on the stored first and second portions of the code. Other embodiments are disclosed and claimed.