Memory Page Migration for HBM Error Avoidance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Advanced memory technologies like HBM pose reliability challenges due to their soldered or bonded nature, making in-field replacement difficult, and there is a need to prevent uncorrectable errors that can lead to system downtime.
Innovation Solution
A method is introduced to monitor correctable errors in memory devices, determining a page migration condition based on error rates, and migrating memory pages to reserved redundancy areas before uncorrectable errors occur, using a mapping table to redirect access and prevent data loss.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If advanced memory technologies (HBM) are used to support high memory bandwidth demands, then memory density and data access rates are improved, but reliability deteriorates due to soldered or bonded nature making in-field replacement difficult
Solution Approach 1:
The system performs preliminary actions by monitoring correctable errors in real-time and proactively migrating memory pages to reserved redundancy areas before uncorrectable errors occur. This preventive approach maintains high reliability without requiring physical replacement of soldered memory devices, thus resolving the contradiction between using advanced memory technologies for high bandwidth and maintaining reliability.
Solution Approach 2:
The patent implements beforehand cushioning by allocating reserved redundancy areas in advance and preparing migration paths for memory pages. When correctable errors are detected, the system has pre-established backup locations ready to receive migrated data, cushioning against potential uncorrectable errors and maintaining system reliability while using high-density HBM for bandwidth performance.
2Reliability
If memory pages are migrated to reserved redundancy areas to avoid uncorrectable errors, then reliability is improved, but device complexity increases due to mapping table management
Solution Approach 1:
The mapping table acts as an intermediary layer between the processor and physical memory locations. It abstracts the complexity of error management and redundancy management, allowing the system to maintain high reliability through page migration while presenting a simple interface to the processor. The mapping table manages the complexity internally, resolving the contradiction between improved reliability and increased device complexity.
3Reliability
If correctable errors are monitored and page migration is performed proactively, then uncorrectable errors are avoided, but loss of time occurs due to migration operations
Solution Approach 1:
The system maintains continuity of useful action by performing page migration in the background without interrupting processor operations. The mapping table is updated to redirect accesses to migrated pages seamlessly, ensuring that the processor continues to execute computations without downtime. This resolves the contradiction between avoiding uncorrectable errors through proactive migration and minimizing loss of time during migration operations.
Data Source
AI summary
Techniques for avoiding uncorrectable errors in a memory device can include detecting a correctable error pattern of a memory page of a memory device, and determining that the correctable error pattern of the memory page satisfies a page migration condition. Upon satisfying the page migration condition, write accesses to the memory page are prevented from reaching a memory controller of the memory device. The contents of the memory page are then migrated to a reserved page, and a mapping table is updated to replace accesses to the memory page with accesses to the reserved page.