PCR*-Tree Node Splitting for Non-Volatile Memory Write Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Non-volatile memory technologies, such as NAND flash and PCM, face challenges in endurance and performance due to high write operations, especially when handling spatial data, as existing data structures like R-trees require frequent node splitting and merging, leading to increased write counts and reduced memory lifespan.
Innovation Solution
The PCR*-tree operating method for non-volatile memory optimizes node splitting and merging by varying the size of leaf nodes relative to non-leaf nodes, allowing for bulk data reinsertion and minimizing writes, thereby reducing the number of writes and improving performance by postponing overflow treatments and optimizing parent node updates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If R-tree data structure is used to handle spatial data, then spatial data storage capability is improved, but the number of write operations increases significantly
Solution Approach 1:
The patent segments the write operations into two distinct phases: a sorting phase that organizes data objects within nodes without writing to memory, and a bulk write phase that writes all changes to memory in a single operation. This segmentation eliminates redundant intermediate writes while maintaining the spatial data structure's organizational capabilities.
Solution Approach 2:
The patent performs preliminary sorting and organization of data objects within nodes before writing to memory. By pre-organizing the data structure in memory and only writing the final sorted result, the system avoids multiple intermediate write operations that would otherwise occur during insertion and splitting operations.
2Adaptability or versatility
If frequent node splitting and merging operations are performed, then data structure adaptability is improved, but memory endurance is reduced
Solution Approach 1:
The patent combines multiple write operations into a single bulk write operation. Instead of writing to memory after each insertion, split, or merge operation, the system accumulates all changes in memory and performs a single consolidated write, thereby reducing the total number of write operations that degrade memory endurance.
Solution Approach 2:
The patent creates a copy of the data structure in memory, performs all necessary sorting and organizational operations on the copy, and then writes the final copied structure to memory in a single operation. This copying approach allows extensive data structure manipulation without corresponding write operations until the final state is achieved.
3Stability of the object's composition
If node sorting is performed before reinsertion, then data organization is improved, but write operations increase
Solution Approach 1:
The patent maintains continuous organization of data objects within nodes during the sorting phase without interrupting to write to memory. The sorting operation continues uninterrupted until completion, after which a single bulk write occurs. This continuous organization approach ensures data stability while minimizing write operations.
Data Source
AI summary
A non-volatile memory includes: at least one parent node; and at least one child node that stores a data object and has a variable size of n multiple of the parent node, n being a natural number, when a target child node to insert at least one new data object is fully filled, wherein the data objects of the target child node are split and reinserted into newly created at least one child node.