L2P Map Virtual Blocks for Low-DRAM Storage Address Translation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing demand for larger L2P maps in storage devices necessitates additional DRAM, which compromises the price competitiveness of storage devices.
Innovation Solution
Implementing an L2P map in an on-demand loading manner using L2P map garbage collection (GC) to maintain the L2P map's currency while reducing DRAM usage, and incorporating a non-volatile memory with an L2P pair region, a volatile memory with an L2V table, and an L2P cache to manage logical to physical page number conversions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If additional DRAM is used for L2P map, then L2P conversion performance is improved, but price competitiveness deteriorates
Solution Approach 1:
The L2P map is segmented into multiple virtual blocks that can be independently managed and loaded. Only the required portions of the L2P map are loaded into DRAM at any given time, rather than loading the entire map, thus reducing DRAM requirements while maintaining conversion performance for active regions.
Solution Approach 2:
The system dynamically loads and unloads virtual blocks of the L2P map based on actual access patterns and requirements. The L2P cache and LPN range map structures enable dynamic adjustment of which L2P entries reside in fast memory versus storage memory, optimizing the balance between performance and memory cost.
2Quantity of substance
If L2P map is reduced in size, then DRAM usage is reduced, but L2P conversion accuracy deteriorates
Solution Approach 1:
An LPN range map structure is introduced as an intermediary index that maps logical page number ranges to specific virtual blocks. This intermediary layer enables efficient location of L2P entries without requiring the entire L2P map to be loaded, maintaining conversion accuracy while reducing memory footprint through hierarchical organization.
Solution Approach 2:
The L2P map organization transitions from a flat single-dimensional structure to a hierarchical multi-dimensional structure with virtual blocks, L2P caches, and LPN range maps. This dimensional organization allows the system to manage large L2P spaces with reduced DRAM by distributing data across multiple hierarchical levels.
3Reliability
If L2P map garbage collection is performed frequently, then L2P map currency is maintained, but random write performance deteriorates
Solution Approach 1:
Garbage collection is performed periodically and selectively on specific virtual blocks rather than continuously across the entire L2P map. The system identifies and collects garbage from virtual blocks that are less frequently accessed or have accumulated more invalid entries, reducing the overall impact on random write performance while maintaining L2P map currency.
Solution Approach 2:
The garbage collection process is applied locally to specific virtual blocks based on their individual characteristics such as garbage accumulation levels and access patterns, rather than uniformly across the entire L2P map. This localized approach concentrates GC efforts where most needed while minimizing disruption to active write operations in other regions.
Data Source
AI summary
A storage device and an operating method of the storage device are provided. The storage device comprises a non-volatile memory including a virtual block including an L2P pair region storing a pair of a logical page number (LPN) and a physical page number (PPN) (LPN-PPN pair) and a data region storing data addressed to the PPN, a volatile memory including an L2V table storing a virtual block number (VBN) corresponding to the LPN, an L2P cache storing the LPN-PPN pair, and an LPN range map storing a portion of the LPN stored in the L2P pair region of the virtual block, and a controller configured to control the non-volatile memory and the volatile memory.