Multilevel Memory Failure Bypass via Proactive Far Memory Offlining
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional multilevel memory systems face reliability and uptime issues due to faults in the intermediate memory hierarchy, where a fault in near memory can lead to errors in multiple far memory locations, reducing platform reliability and server availability.
Innovation Solution
Implementing multilevel memory error management techniques that proactively offline far memory locations mapped to a faulty near memory location and maintain a faulty near memory location list to prevent future accesses, thereby enhancing platform reliability and server uptime.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional multilevel memory systems are used without proactive error management, then device complexity is reduced, but system reliability deteriorates due to faults in near memory causing errors in multiple far memory locations
Solution Approach 1:
The patent applies preliminary action by proactively offlining far memory locations that are mapped to a faulty near memory location before actual errors occur. When a fault is detected in near memory, the system preemptively identifies and takes offline all far memory locations that map to the faulty near memory location, preventing potential errors before they happen. This is implemented through firmware or operating system logic that monitors near memory health and automatically offlines corresponding far memory locations in advance.
2Reliability
If proactive error management is implemented to offline far memory locations, then system reliability is improved, but device complexity increases due to additional error management mechanisms
Solution Approach 1:
The patent uses an intermediary approach by introducing a faulty near memory location list that acts as a mediator between near memory faults and far memory management. This list stores information about faulty near memory locations and serves as an intermediary data structure that enables the system to identify and offline corresponding far memory locations without requiring complex direct mapping logic. The intermediary list simplifies the error management structure by providing a centralized repository for fault information.
3Reliability
If all far memory locations mapped to a faulty near memory location are offlined, then reliability is improved by preventing future errors, but loss of time increases due to memory access interruptions
Solution Approach 1:
The patent applies the taking out principle by extracting and isolating only the specific far memory locations that are mapped to faulty near memory locations, rather than offlining entire memory banks or systems. This selective extraction approach minimizes the impact on overall system operation by taking offline only the specific memory locations that would cause errors, while leaving the rest of the memory system operational and accessible.
Data Source
AI summary
Multilevel memory error management techniques can improve system performance, availability, and reliability by preventing future accesses to faulty near memory locations. According to examples described herein, multilevel memory error management techniques enable proactively offlining far memory locations mapped to a faulty near memory location before additional faults are encountered, and/or maintaining a faulty near memory location list to enable bypassing the faulty near memory location to prevent future errors.


