Auxiliary Ethernet Network for PCIe Switch Initialization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In high-computing communication systems using PCIe switches, initialisation of processors is challenging due to asynchronous start-up, inability to determine the state of remote processors, and rigid configuration requirements, leading to potential errors and resource underutilisation.
Innovation Solution
A communication system with a main PCIe network and an auxiliary ETHERNET-type network, where the auxiliary network is used for initialisation, allowing dynamic configuration and adaptation to changes without oversizing components, by broadcasting processor information and using a master processor to configure address translations and routing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If static reservation of all communication areas and static programming of all address translations is implemented, then network initialisation reliability is improved, but network adaptability deteriorates
Solution Approach 1:
The patent segments the communication network into two distinct networks: a first communication network (PCIe) for high-performance data transmission and a second communication network (Ethernet) for initialisation and configuration. This segmentation allows each network to have specialised functions - the Ethernet network handles reliable initialisation and dynamic configuration, while the PCIe network focuses on high-speed communication, resolving the contradiction between reliability and adaptability.
Solution Approach 2:
The second communication network (Ethernet) acts as an intermediary that facilitates the initialisation and configuration of the first communication network (PCIe). Through this intermediary, processors can exchange initialisation messages, establish address translations, and configure routing before high-performance communication begins, enabling both reliable initialisation and subsequent adaptability.
2Adaptability or versatility
If communication areas are oversized to allow modifications during operation, then network adaptability is improved, but resource utilisation deteriorates
Solution Approach 1:
The patent implements dynamic address translation and routing configuration through the exchange of initialisation messages in the Ethernet network. Address translations and routing tables are built dynamically based on actual processor needs and network state, allowing the system to adapt to modifications during operation without requiring oversized pre-configured communication areas, thus maintaining both adaptability and efficient resource utilisation.
3Adaptability or versatility
If address translations are used for communication, then communication flexibility is improved, but configuration complexity deteriorates
Solution Approach 1:
The patent implements a self-service mechanism where processors automatically exchange initialisation messages in the Ethernet network to discover each other's addresses and capabilities. Each processor autonomously builds its address translation table and routing information based on received messages, eliminating the need for complex external configuration and reducing configuration complexity while maintaining communication flexibility.
Data Source
AI summary
The present invention relates to a communication system comprising a plurality of processors and at least one switch, referred to as a main switch, connecting the processors into a main communication network;the system being characterised in that it further comprises at least one other switch, called an auxiliary switch, connecting the processors in an auxiliary communication network, and in that the auxiliary communication network is intended to be used by the processors to initialise the main communication network.