Cache Metadata Management in Hybrid Disk Drives
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional cache management schemes in hybrid disk drives are inefficient due to the large size of the in-DRAM hash table, which exceeds the limited DRAM capacity, leading to impractical performance and high power consumption.
Innovation Solution
A data storage device and method that uses a metadata table with a one-to-one mapping between data blocks and cache blocks, where each data block's cache block address and status are stored in a reserved area, allowing direct access without computation, and a bitmap is used to quickly identify the status of cache blocks, reducing the DRAM required for metadata storage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If a conventional set associative hash table is used to store cache metadata, then the cache management can be performed, but the DRAM capacity required becomes excessively large exceeding the limited DRAM capacity
Solution Approach 1:
The patent extracts the cache metadata storage from the DRAM and relocates it to the NV cache memory. The DRAM only retains a small bitmap for status indication, while the comprehensive metadata (LBA-to-CBA mappings, block addresses, status information) is stored in the NV cache. This extraction resolves the contradiction by eliminating the need for large DRAM capacity while maintaining cache management functionality.
Solution Approach 2:
The patent introduces an intermediary structure (metadata table in NV cache) that mediates between the DRAM and the actual data storage. The NV cache acts as an intermediary to store the bulk of metadata, allowing the DRAM to function with minimal capacity requirements. This intermediary approach enables efficient cache management without requiring excessive DRAM.
2Productivity
If the DRAM size is increased to store more cache metadata, then the cache management performance improves, but the power consumption increases
Solution Approach 1:
The patent extracts the metadata storage function from the DRAM and relocates it to the NV cache. By removing the heavy metadata storage burden from the DRAM, the system achieves good cache management performance without requiring large DRAM capacity, thereby avoiding the associated power consumption increase. The DRAM only needs to maintain a small bitmap for status information.
3Productivity
If a hash table with linear search is used for cache metadata, then the cache blocks can be managed, but the query and update operations become inefficient
Solution Approach 1:
The patent replaces the mechanical linear search process with a direct addressing mechanism. Instead of using a hash table that requires computation and linear search through the metadata, the system uses a direct mapping approach where metadata for each data block is stored at a predetermined location in the NV cache. This substitution eliminates the time-consuming linear search and computation steps, significantly improving query and update efficiency.
Data Source
AI summary
A data storage device includes a data storage medium, a cache, and a cache control memory. The data storage medium has M data blocks. M is an integer greater than 1. The cache includes N cache blocks having N cache block addresses, respectively. N is an integer greater than 1. The cache control memory includes M memory elements corresponding to the M data blocks, respectively. The cache control memory is configured to, in response to a request to cache data of one of the M data blocks: (a) write the data from the one of the M data blocks to one of the N cache blocks; and (b) write, in the one of the M memory elements corresponding to the one of the M data blocks, one of the N cache block addresses corresponding to the one of the N cache blocks where the data is written.