Scope-Selective Coherency Broadcast in Multinode Data Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional cache coherent data processing systems face scalability issues due to the need for global broadcasts of coherency messages, leading to increased access latency and bandwidth requirements as system size increases, which can result in inefficient communication and higher costs.
Innovation Solution
Implementing non-overlapping coherency domains within the cache coherent data processing system, allowing a master to select a scope for an initial broadcast of interconnect operations, such as a local or remote scope, to limit communication and reduce unnecessary global operations, thereby conserving bandwidth and reducing operational latency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If global broadcast of coherency messages is employed to maintain cache coherency, then cache coherency is ensured across all processing units, but system scalability deteriorates and access latency increases as system size increases
Solution Approach 1:
The system divides the multiprocessor system into multiple coherency domains, where each domain is a subset of processing units that share coherency information. This segmentation allows coherency messages to be restricted to specific domains rather than being globally broadcast to all processing units, thereby improving scalability while maintaining coherency within each domain.
Solution Approach 2:
The patent implements domain-specific coherency management where each coherency domain has its own coherency state information and message handling. Coherency messages are routed to specific domains based on the target address, ensuring that only relevant processing units receive and process the messages, thus reducing unnecessary traffic and improving system scalability.
2Reliability
If global broadcast of coherency messages is used, then cache coherency is maintained across all processing units, but bandwidth requirements increase
Solution Approach 1:
By segmenting the system into coherency domains, the patent reduces the scope of coherency message broadcasts from global to domain-specific. This segmentation limits the number of processing units that need to receive and process each coherency message, thereby reducing the overall bandwidth consumption on the system interconnect.
Solution Approach 2:
The patent applies partial action by broadcasting coherency messages only to the necessary coherency domains rather than to all processing units in the system. This partial broadcasting approach reduces unnecessary message transmission and minimizes bandwidth consumption while still ensuring coherency where needed.
3Reliability
If global broadcast of coherency messages is implemented, then cache coherency is ensured, but access latency increases
Solution Approach 1:
The patent segments the coherency message distribution into domain-specific broadcasts rather than global broadcasts. This segmentation reduces the time required for message delivery since fewer processing units need to be notified, thereby reducing access latency while maintaining coherency within each domain.
Solution Approach 2:
The system performs preliminary actions by pre-establishing coherency domains and routing information before coherency messages are generated. This preliminary organization allows for efficient message routing and reduces the time required to distribute coherency information, thereby reducing access latency.
Data Source
AI summary
A cache coherent data processing system includes at least non-overlapping first, second, and third coherency domains. A master in the first coherency domain of the cache coherent data processing system selects a scope of an initial broadcast of an interconnect operation from among a set of scopes including (1) a remote scope including both the first coherency domain and the second coherency domain, but excluding the third coherency domain that is a peer of the first coherency domain, and (2) a local scope including only the first coherency domain. The master then performs an initial broadcast of the interconnect operation within the cache coherent data processing system utilizing the selected scope, where performing the initial broadcast includes the master initiating broadcast of the interconnect operation within the first coherency domain.