Fault-Tolerant Management Fabric for Multipartition Servers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing rack systems face challenges in managing and monitoring high availability servers due to complex and costly hardware requirements for management subsystems, which increase complexity and costs while compromising on space and processing resources.
Innovation Solution
A redundant, fault-tolerant management fabric is implemented, allowing for virtualization of resources and control functions, reducing the need for dedicated hardware on system boards by using a separate processor and memory, and providing multiple paths for network redundancy, ensuring high availability and low complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a management subsystem with dedicated hardware is implemented to monitor and control server processors, then system reliability and management capability are improved, but device complexity and cost increase
Solution Approach 1:
The patent extracts the management subsystem functionality from dedicated hardware on system boards and consolidates it into a separate, shared management processor. This management processor is physically separated from the server processing resources it manages, allowing management functions to be performed without adding complexity to the core server hardware while maintaining full management capability.
Solution Approach 2:
The shared management processor serves multiple system boards simultaneously, providing universal management capability across the entire server system. A single management processor can monitor and control multiple processors across different system boards, eliminating the need for dedicated management hardware on each board and reducing overall system complexity.
2Reliability
If dedicated management hardware is installed on each system board to ensure fault tolerance, then system reliability is improved, but the space and cost resources are consumed
Solution Approach 1:
The patent merges the management functions of multiple system boards into a single shared management processor. Instead of having separate management hardware on each system board, the management capabilities are combined and consolidated into one centralized processor that serves the entire system, reducing hardware resource consumption while maintaining management functionality.
Solution Approach 2:
The shared management processor provides universal service to multiple system boards, allowing a single hardware resource to perform management functions across the entire system. This multi-functional approach eliminates the need for redundant management hardware on each board while maintaining fault tolerance through the distributed architecture.
3Reliability
If redundant management paths are implemented to ensure high availability, then system reliability is improved, but device complexity increases
Solution Approach 1:
The patent introduces a network fabric as an intermediary layer between the shared management processor and the system boards. This network fabric provides redundant communication paths without requiring complex direct connections between the management processor and each system board. The network fabric absorbs the complexity of redundancy implementation, allowing the management architecture to remain simple while achieving high availability.
Data Source
AI summary
Redundant, fault-tolerant management fabric for multipartition servers are disclosed. In an exemplary embodiment, a method comprises connecting a plurality of rack system components to a first network segment, the connection including at least two physical links sharing a single network address. The method also comprises monitoring communications paths in the first network segment. The method also comprises switching communications from the first network segment to a failover network segment if there is a failure in any of the communications paths in the first network segment.