Multi-Processor Cluster Transaction Routing via Protocol Engines
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Multi-processor systems with point-to-point communication infrastructure are limited by a fixed node ID address space, leading to a restricted number of interconnected nodes and a single level of address space mapping, which also restricts processing throughput due to the capacity of the protocol engine.
Innovation Solution
Implementing multiple protocol engines in each cluster to process transactions and using a hierarchical mapping mechanism that allows interconnection of multiple clusters, enabling a shared memory address space and increased processing throughput by mapping transaction packets based on destination information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Power
If a point-to-point communication infrastructure with fixed node ID address space is used, then the system achieves increased information transmission capacity, but the number of interconnected nodes is limited
Solution Approach 1:
The system is divided into multiple processor clusters, each with its own interconnection controller and protocol engines. This segmentation allows each cluster to operate independently with its own node ID space, enabling the overall system to scale beyond the fixed address space limitation of a single cluster while maintaining high transmission capacity within each cluster.
Solution Approach 2:
The patent introduces a hierarchical address mapping mechanism that adds a cluster-level dimension to the traditional flat node ID address space. By mapping global addresses to cluster addresses and then to node addresses, the system extends the addressable node space from a single fixed level to a multi-level hierarchy, enabling connection of multiple clusters without increasing the base node ID space.
2Device complexity
If a flat address space mapping is used, then the system structure is simple, but only a single level of mapping is allowed
Solution Approach 1:
The patent transforms the flat single-level address mapping into a multi-level hierarchical mapping structure. The global address space is divided into cluster address spaces, which are further divided into node address spaces. This dimensional transformation allows the system to support multiple levels of mapping while maintaining manageable complexity at each level through standardized mapping mechanisms.
Solution Approach 2:
The address mapping structure implements a nested hierarchy where global addresses contain cluster address information, which in turn contain node address information. Each level of the hierarchy is nested within the previous level, allowing the system to maintain a structured organization of multiple mapping levels while enabling efficient address translation through staged lookup processes.
3Device complexity
If a single protocol engine is used to process transactions, then the system structure is simple, but the processing throughput is limited
Solution Approach 1:
The interconnection controller is divided into multiple protocol engines, each capable of independently processing transactions. This segmentation of the protocol processing function allows parallel handling of multiple transactions simultaneously, significantly increasing throughput. Each protocol engine operates as an independent processing unit, eliminating the bottleneck of a single sequential processor.
Solution Approach 2:
Multiple protocol engines are combined within a single interconnection controller, creating a unified controller that leverages the collective processing power of all engines. The engines share common resources such as the crossbar switch and memory interfaces while maintaining independent transaction processing capabilities, achieving both high throughput and resource efficiency.
Data Source
AI summary
A multi-processor computer system is described in which transaction processing in each cluster of processors is distributed among multiple protocol engines. Each cluster includes a plurality of local nodes and an interconnection controller interconnected by a local point-to-point architecture. The interconnection controller in each cluster comprises a plurality of protocol engines for processing transactions. Transactions are distributed among the protocol engines using destination information associated with the transactions.