Hardware Memory Migration Engine for Server Reliability

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current server memory controllers rely on software-based mirror/memory migration, which is inefficient and can lead to system failure due to the potential consumption of poisoned data during the migration process, especially in modern architectures with multi-core processors and multiple cache levels.

Innovation Solution

Implementing hardware-based memory migration and mirroring techniques that allow for transparent data movement between physical memory addresses, using a master and slave node architecture with Patrol Scrub Engines and memory coherency agents to manage error detection and fail-over, thereby reducing the need for software intervention and minimizing the risk of poisoned data consumption.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If software-based memory migration is used, then memory fail-over can be implemented, but system overhead increases and migration throughput decreases

Engineering Contradiction:

Improvememory fail-over capabilityVSAvoidmigration throughput

Core Design Contradiction:

ReliabilityVSProductivity

Solution Approach 1:

The patent replaces software-based memory migration with a hardware-based implementation using memory controllers and dedicated migration circuitry. The memory controller includes a migration engine that performs data movement at the hardware level, eliminating the overhead of software intervention and achieving higher migration throughput while maintaining fail-over reliability.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent introduces a dedicated migration engine as an intermediary component within the memory controller. This migration engine acts as a specialized mediator between memory controllers, handling the data transfer operations during migration and fail-over scenarios, thereby improving efficiency and reducing system overhead.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Reliability

If software-based memory migration is used, then memory fail-over can be implemented, but system overhead increases

Engineering Contradiction:

Improvememory fail-over capabilityVSAvoidsystem overhead

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent enables the memory controller to perform migration operations autonomously through its integrated migration engine. The system serves itself by handling migration and fail-over operations at the hardware level without requiring external software management, thereby reducing system overhead and complexity.

Inventive Principle:

Principle #25Self-service

3Reliability

If software-based memory migration is used, then memory can be migrated, but poisoned data may be consumed during migration

Engineering Contradiction:

Improvememory migration capabilityVSAvoiddata integrity

Core Design Contradiction:

ReliabilityVSLoss of information

Solution Approach 1:

The patent implements preliminary error detection and data validation mechanisms before migration operations commence. The migration engine checks for poisoned or corrupted data and prevents its migration, ensuring data integrity is maintained throughout the migration process and eliminating the risk of consuming poisoned data.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS10061534B2Hardware based memory migration and resilvering

Publication Date: 2018.08.28 INTEL CORP

AI summary

Method, apparatus and systems for performing hardware-based memory migration and copy operations. Under the method, a first portion of memory in a computer system accessed via a first memory controller is migrated or copied to a second portion of memory accessed via a second memory controller using a hardware-based scheme that is implemented independent of and transparent to software running on a computer system. The memory migration and/or copy operations can be used to initialize a memory mirror configuration under which data in first and second portions of memory are mirrored, and to perform memory migration operations in which data in a first portion of memory is migrated to a second portion of memory under the control of hardware in a manner in which the memory migration can be performed during run-time without a significant reduction in performance. In addition, poison data indicating failed cache lines may be migrated or copied such that data corresponding to migrated or copied poisoned cache lines are not used.