Nested Cache Coherency Fabric for Multi-Node Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Modern multi-processor computer systems face challenges in maintaining cache coherency across multiple processors, particularly in large SMP systems, where existing protocols are not scalable and efficient in managing cache coherency between nodes, leading to inconsistencies and performance issues.
Innovation Solution
A multi-tiered cache coherency fabric is introduced, comprising local and global cache coherency fabrics, where local coherency is managed within nodes and global coherency between nodes, with logic determining the appropriate coherency operation based on the cache state and request origin, interlocking concurrent requests, and suppressing unnecessary operations to maintain cache coherency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a single global cache coherency protocol is used across all nodes, then cache coherency can be maintained system-wide, but system complexity increases and performance degrades due to excessive node-to-node communication
Solution Approach 1:
The patent divides the cache coherency system into two independent segments: a local cache coherency fabric within each node and a global cache coherency fabric between nodes. Each fabric operates independently with its own protocol, allowing the local fabric to handle intra-node coherency without involving other nodes, thereby reducing overall system complexity while maintaining coherency reliability.
Solution Approach 2:
The patent introduces a hierarchical dimension to cache coherency management by organizing coherency operations at two levels: local (intra-node) and global (inter-node). This dimensional organization allows coherency to be maintained at each level independently, reducing the need for system-wide communication and simplifying the overall protocol structure.
2Reliability
If all cache coherency requests are handled globally across all nodes, then coherency is maintained, but latency increases due to additional node-to-node hops
Solution Approach 1:
By segmenting coherency requests into local and global handling domains, the patent allows local requests to be resolved within the node without traversing the global network, thereby reducing latency while maintaining coherency through the global fabric only when necessary.
Solution Approach 2:
The local cache coherency fabric performs preliminary coherency actions within the node before global fabric involvement is needed. This preliminary handling of local requests reduces the number of global node-to-node hops required, thereby reducing latency while ensuring coherency is maintained.
3Reliability
If unnecessary cache coherency operations are performed, then coherency is maintained, but system performance decreases due to redundant operations
Solution Approach 1:
The patent applies partial action by having the local cache coherency fabric handle only the portion of coherency requests that are local in nature, without performing excessive global fabric operations. This selective handling reduces redundant operations while maintaining necessary coherency, thereby improving system performance.
Solution Approach 2:
By segmenting coherency operations into local and global scopes, the patent prevents unnecessary global fabric involvement for local requests, reducing redundant operations and improving performance while maintaining coherency where required.
Data Source
AI summary
A computer system comprising multiple nodes, each node comprising a plurality of processors and a local cache hierarchy, suppresses local cache coherency of a node operations or global cache coherency operations between nodes based on the coherency request being a global or local request, and the state of the cache line at the node.