Nonvolatile Memory Wear Management via In-Unit Metadata
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wear-leveling techniques for nonvolatile memory require complex firmware and large indirection tables in DRAM to prevent over-cycling, which is inefficient and resource-intensive.
Innovation Solution
Implementing targeted wear management by storing metadata in each storage unit to indicate its state, allowing for efficient management of highly utilized areas without the need for large indirection tables, using a phase-change memory array with metadata that includes state information, cycle count, and error correction codes to identify and manage proxy and non-proxy states.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If wear-leveling is implemented using large indirection tables in DRAM, then over-cycling prevention is effective, but firmware complexity increases considerably
Solution Approach 1:
The patent divides the wear management approach into segments: instead of using a single large indirection table for all memory management, it implements targeted wear management that segments the problem into identifying only highly utilized areas and managing those specifically. This is achieved through storing metadata in each storage unit that indicates whether it is in a proxy state or non-proxy state, allowing wear management to focus only where needed rather than managing all memory uniformly.
Solution Approach 2:
The patent extracts the essential wear management function from the complex indirection table system. Instead of relying on large DRAM-based indirection tables, it extracts the critical information needed for wear management into compact metadata stored directly in each storage unit. This metadata contains only the necessary state information (proxy or non-proxy state), removing the need for large external tables while retaining wear management capability.
2Reliability
If large indirection tables are used in DRAM for logical-to-physical address mapping, then wear-leveling is effective, but memory resources are consumed inefficiently
Solution Approach 1:
The patent extracts the essential wear management information from large DRAM-based indirection tables and places it directly in the storage units as compact metadata. This extraction eliminates the need for large DRAM resources dedicated to indirection tables, as only essential state information (proxy or non-proxy state) is stored in each storage unit's metadata, significantly reducing DRAM consumption.
Solution Approach 2:
The patent implements self-service wear management where each storage unit contains its own metadata indicating its state (proxy or non-proxy). This allows the storage unit to effectively manage its own wear characteristics without requiring external DRAM resources for tracking, as the necessary information is self-contained within each storage unit's metadata structure.
Data Source
AI summary
Embodiments describe methods, apparatus, and system configurations for providing targeted wear management in nonvolatile memory. Specifically, embodiments may include a memory controller to receive a memory access request directed to a storage unit of the nonvolatile memory. The memory controller may access metadata in the storage unit and determine whether to perform a memory access in accordance with the memory access request based on the state information. Other embodiments may be described or claimed.