Retired Page Table Tracking for Persistent Memory Fault Isolation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In conventional computer systems, memory errors in retired physical memory are not persisted across boot cycles, leading to potential further faults and compromised performance due to the loss of error data after reboot, especially in environments like data centers where faulty memory can be moved without proper tracking.
Innovation Solution
A method involving a hardware processor that stores a retired page table in read-only memory, allowing it to persist memory error data across boot cycles by comparing SoC records and retiring faulty physical pages, and publishes this data to the operating system and other systems for tracking and managing memory errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If memory error data is stored in volatile memory, then the system can quickly access and process error data during operation, but the error data is lost after reboot causing faulty memory to be reused
Solution Approach 1:
The patent divides the error data storage function into two separate components: volatile memory for fast access during operation and non-volatile memory for persistent storage across reboots. This segmentation allows the system to maintain both quick error data retrieval and data persistence, resolving the contradiction between speed and information retention.
Solution Approach 2:
The patent introduces non-volatile memory as an intermediary between the volatile memory and the error tracking system. This intermediary component captures and preserves error data that would otherwise be lost, enabling the system to maintain fast operational access while preventing information loss during system restarts.
2Reliability
If the system retires faulty memory pages, then system reliability improves by preventing further faults, but available memory capacity decreases
Solution Approach 1:
The patent implements preliminary retirement of memory pages by marking them as retired before they are actually needed. The retired page table pre-identifies and marks faulty memory pages during system initialization or when errors are detected, preventing their future use. This preliminary action ensures reliability by proactively excluding bad memory while the system can still utilize healthy memory capacity.
Solution Approach 2:
The patent creates a retired page table that serves as a copy or map of the actual memory state, tracking which pages are faulty without physically removing them. This virtual copying approach allows the system to maintain reliability through accurate tracking while preserving the physical memory structure and maximizing usable capacity.
3Measurement precision
If the retired page table is updated frequently, then the error tracking remains current and accurate, but the complexity of maintaining the page table increases
Solution Approach 1:
The patent implements periodic updates to the retired page table rather than continuous updates. The system updates the page table at specific intervals or triggered by specific events such as memory errors or system state changes. This periodic approach maintains accurate error tracking while reducing the overall complexity and overhead of page table maintenance compared to continuous updates.
Data Source
AI summary
Defect tracking within a computing environment includes storing a retired page table within a read only memory (ROM) of a computing node. A hardware processor of the computing node detects one or more System-on-Chips (SoCs) within the computing node. The hardware processor reads the retired page table from the read only memory. The hardware processor compares records of the retired page table with the one or more SoCs within the computing node. For each record of the retired page table that matches the one or more SoCs within the computing node, one or more physical pages of a volatile memory of the one or more SoCs are retired based on each record.