PCI Express Bridge I/O Sharing via VLAN Identifier Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for sharing I/O equipment among multiple CPUs are complex and lack expandability, with difficulties in group management and setting redundancy, leading to synchronization issues and limited scalability.
Innovation Solution
A system where multiple CPUs are connected through upstream and downstream PCI Express bridges with identifier setting and filtering units, allowing for easy sharing and management of I/O equipment by setting the same identifier and communicating using frames with added identifiers, thereby simplifying network management and increasing redundancy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If PCI Express switch with upstream and downstream PCI-PCI bridges is used, then I/O equipment can be connected to CPU, but only tree-like arrangement is permitted and sharing among multiple CPUs is not possible
Solution Approach 1:
The system segments the connection architecture into multiple independent PCI Express switches, each capable of connecting to multiple CPUs. This allows I/O equipment to be shared across different CPU clusters while maintaining manageable connection units, resolving the limitation of tree-like arrangements and enabling flexible sharing without excessive complexity.
Solution Approach 2:
Each PCI Express switch is designed with multi-functionality to serve multiple CPUs and multiple I/O equipment simultaneously. The switch can dynamically allocate and manage connections between different CPU-I/O pairs, enabling universal access and sharing capabilities while maintaining a standardized connection interface that reduces overall system complexity.
2Adaptability or versatility
If ASI network with fabric manager is used to share I/O equipment among multiple CPUs, then sharing is possible, but group management becomes complex and expandability is limited
Solution Approach 1:
The system divides the network into multiple independent PCI Express switches that can operate autonomously. Each switch manages its own connections locally without requiring a centralized fabric manager, thereby simplifying network management while enabling I/O sharing among multiple CPUs through the distributed switch architecture.
Solution Approach 2:
Each PCI Express switch performs self-service by autonomously managing its own connection routing and resource allocation. The switches independently handle connection establishment and resource sharing without external coordination, eliminating the need for complex centralized management and improving system expandability.
3Reliability
If multiple I/O equipment are connected through single PCI Express switch, then connection is established, but redundancy setting is difficult and synchronization issues occur
Solution Approach 1:
The system segments the connection path by introducing multiple PCI Express switches in parallel, allowing I/O equipment to be connected through multiple independent paths. This segmentation enables redundancy configuration where the same I/O equipment can be accessed through different switch routes, improving reliability while maintaining manageable connection complexity.
Solution Approach 2:
The system changes the connection parameter from single-path to multi-path topology by configuring multiple PCI Express switches. This parameter change enables redundancy settings where alternative connection paths are available, and synchronization is maintained through standardized communication protocols across the parallel paths.
Data Source
AI summary
An I/O equipment sharing system includes CPUs, a plurality of route complexes coupled to the CPUs, upstream PCI Express-bridges coupled to the route complexes, downstream PCI Express-bridges coupled to the upstream PCI Express-bridges through a network, and I/O equipment coupled to the downstream PCI Express-bridges. In the above configuration, the I/O equipment are shared between the CPUs using the identifiers of the network (for example, Ethernet VLAN IDs), the identifiers are set so that they do not overlap between the respective CPUs and necessary I/O equipment is set to a set identifier. Further, an identifier is set to a plurality of the same I/O equipment required by the respective CPUs.