CPU Hot-Remove and Hot-Add Topology Management

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

In multi-CPU interconnection systems, existing methods require powering off the server to replace a faulty CPU, leading to system service interruptions and affecting continuous operation.

Innovation Solution

The CPU hot-remove and hot-add methods allow for online addition or removal of CPUs without powering off the system, maintaining a stable topology and ensuring continuous operation by identifying and removing corresponding CPUs to maintain a stable CPU structure.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If multiple CPUs are interconnected to improve processing performance, then system performance is improved, but system reliability deteriorates because the entire system may crash if any CPU is faulty

Engineering Contradiction:
Improveprocessing performanceVSAvoidsystem stability
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent segments the CPU interconnection system into independent functional modules, where each CPU can be managed and replaced independently. The CPU topology is divided into replaceable units that can be hot-swapped without affecting the entire system, thus maintaining reliability while preserving high-performance interconnection capabilities.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent implements preparatory measures by establishing a stable CPU topology structure before failures occur. The system pre-configures replacement protocols and topology validation mechanisms that are ready to activate immediately when a CPU failure is detected, preventing system crashes and maintaining continuous operation.

Inventive Principle:
Principle #11Beforehand cushioning (Prior cushioning)

2Ease of repair

If the server is powered off to replace a faulty CPU, then the CPU can be changed, but system service is interrupted affecting continuous operation

Engineering Contradiction:
ImproveCPU replacement capabilityVSAvoidservice interruption time
Core Design Contradiction:
Ease of repairVSLoss of time

Solution Approach 1:

The patent performs preliminary actions by validating the CPU topology and preparing replacement protocols before the actual CPU replacement occurs. The system checks topology stability, identifies suitable replacement CPUs, and pre-configures interconnection paths, enabling hot-swapping without service interruption.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent implements dynamic CPU replacement capabilities where the system can adapt its topology and interconnection paths in real-time during CPU replacement. The CPU topology is designed to be dynamically reconfigurable, allowing live swapping of CPUs while maintaining system operation through automated topology adjustment and resource reallocation.

Inventive Principle:
Principle #15Dynamics

3Adaptability or versatility

If CPUs are added or removed from the system, then system capacity can be adjusted, but the CPU topology may become unstable

Engineering Contradiction:
Improvesystem capacity flexibilityVSAvoidCPU topology stability
Core Design Contradiction:
Adaptability or versatilityVSStability of the object's composition

Solution Approach 1:

The patent implements feedback mechanisms that continuously monitor CPU topology stability after addition or removal operations. The system validates topology integrity, detects instability conditions, and automatically triggers corrective actions such as topology reconfiguration or CPU reassignment, ensuring that capacity adjustments do not compromise system stability.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The patent utilizes parameter changes in CPU topology configuration to maintain stability during capacity adjustments. By carefully controlling topology parameters such as interconnection patterns, resource allocation ratios, and CPU grouping configurations, the system enables flexible capacity changes while preserving topological stability and preventing system crashes.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3767470B1Central processing unit CPU hot-remove method and apparatus, and central processing unit CPU hot-add method and apparatus
Publication Date: 2022.02.23 HUAWEI TECH CO LTD
  • EP3767470B1 patent drawingFigure 1~2
  • EP3767470B1 patent drawingFigure 3
  • EP3767470B1 patent drawingFigure 4

AI summary

A central processing unit CPU hot-remove method and apparatus, and a central processing unit CPU hot-add method and apparatus are provided. The method is applicable to a server that has a non-full mesh first CPU topology, and the method includes the following steps: A controller determines a first CPU from multiple CPUs (S710), where the first CPU is a CPU that is faulty or that needs to be removed according to first indication information, and the first indication information is from the first CPU topology or a user interface. The controller determines at least one second CPU from the multiple CPUs, where the at least one second CPU and the first CPU meet a preset condition (S720). The controller sends second indication information to the first CPU topology (S730), where after the first CPU topology receives the second indication information, the first CPU and the at least one second CPU are removed, so as to obtain a second CPU topology and run the second CPU topology. According to the foregoing method, in-service CPU removal can be achieved, and during a CPU removing process and after the CPU is removed, a system can work normally, thereby improving user experience.