Granular Health Visibility for Persistent Memory Ranks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current information handling systems lack granular health visibility for non-volatile memory modules, as they only provide a single health bit for the entire module, failing to account for varying failure rates across different portions of the media, such as blocks or ranks, which can lead to unreliable save operations.
Innovation Solution
Incorporating health registers accessible to the host system for each rank of volatile memory within persistent memory modules, which store information on the reliability of save operations to associated non-volatile memory, allowing for granular health monitoring and data allocation optimization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single health bit is used for the entire NVDIMM module, then the device complexity is reduced, but the measurement precision of media health is insufficient
Solution Approach 1:
The patent divides the NVDIMM module into multiple ranks, with each rank having its own health bit in separate health registers. This segmentation allows the host system to identify which specific rank has failed, providing granular health visibility while maintaining manageable device complexity through structured organization.
Solution Approach 2:
The patent introduces a new dimension of health monitoring by creating separate health registers for each rank rather than using a single aggregate health bit. This dimensional expansion transforms the health monitoring from a binary module-level indicator to a multi-granularity system that provides both overall module health and specific rank-level diagnostics.
2Measurement precision
If granular health information per rank is provided, then the measurement precision of media health is improved, but the device complexity increases
Solution Approach 1:
The health monitoring structure is segmented into multiple health registers, one for each rank of the NVDIMM module. Each register contains a health bit specific to that rank, enabling precise identification of failed ranks while organizing the complexity into manageable, standardized units that follow memory hierarchy conventions.
Solution Approach 2:
Each rank's health status is independently tracked and reported through its dedicated health register, allowing the memory controller and host system to self-diagnose rank-specific failures without requiring complex external analysis or additional monitoring infrastructure beyond the standard memory interface.
3Ease of operation
If a single health bit for the entire module is used, then the ease of operation is maintained, but the reliability of save operations is reduced
Solution Approach 1:
The health information is segmented into rank-specific bits within health registers, allowing the host system to maintain simple read operations while gaining the ability to identify exactly which rank has failed. This enables reliable save operations by preventing writes to failed ranks while keeping the interface straightforward.
Solution Approach 2:
The health registers provide continuous feedback to the host system about the status of each rank, enabling the system to adaptively manage save operations by routing data only to healthy ranks. This feedback mechanism maintains ease of operation through standard memory interfaces while dramatically improving save operation reliability.
4Reliability
If rank-level health registers are implemented, then the reliability of save operations is improved, but the loss of information increases due to more complex health monitoring requirements
Solution Approach 1:
Health information is segmented into compact, rank-specific bits within standard memory-mapped registers, minimizing the overhead of health monitoring. Each health bit consumes only one bit of storage per rank, and the registers are accessed through existing memory interface protocols, reducing the overall information overhead while maximizing reliability.
Solution Approach 2:
The health information is represented as simple binary status bits within standard register structures, changing the parameter representation to the most efficient form. This uses minimal storage space (one bit per rank) and leverages existing memory interface capabilities, reducing information overhead while providing comprehensive health visibility for reliable operation management.
Data Source
AI summary
In accordance with embodiments of the present disclosure, an information handling system may include a processor and a memory system communicatively coupled to the processor. The memory system may include one or more persistent memory modules, each of the one or more persistent memory modules comprising a volatile memory and a non-volatile memory and one or more health registers accessible to a host system executing on the processor, the health registers storing health information indicating, for each of a plurality of ranks of the volatile memory, whether the memory system can reliably perform a save operation to a portion of non-volatile memory mapped to volatile memory of the rank.