Distributed Arbitration for Shared Data Path Latency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional system-on-chip (SOC) architectures face inefficiencies in arbitrating access to shared resources due to increased latency and reduced performance when dealing with multiple data sources, particularly as sources farther from the receiver experience greater transmission delays, leading to unfair bandwidth distribution.
Innovation Solution
A data pipeline with multiple data arbiters, each equipped with a multiplexer, register, and controller, selectively passes upstream or local data packets based on stored indications of packet sources, ensuring fair and efficient access by distributing arbitration along the pipeline, using a combination of multiplexers, registers, and controllers to manage packet flow and maintain a count of packets from each source.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single centralized arbiter is used to manage all data sources, then the arbitration logic is simplified and device complexity is reduced, but sources farther from the receiver experience greater transmission delays leading to increased latency and unfair bandwidth distribution
Solution Approach 1:
The patent divides the centralized arbitration function into multiple distributed arbiters positioned at different points along the data pipeline. Each arbiter manages a subset of data sources locally, segmenting the arbitration process to reduce transmission distance and latency for each source while maintaining manageable complexity at each node.
Solution Approach 2:
The patent introduces a spatial dimension to arbitration by distributing arbiters along the data pipeline rather than using a single centralized point. This dimensional transformation allows sources to access the shared resource through nearer arbitration points, reducing transmission latency while the hierarchical structure maintains overall system manageability.
2Productivity
If multiple distributed arbiters are deployed along the data pipeline, then transmission latency is reduced and bandwidth distribution fairness is improved, but device complexity and the number of arbitration components increase
Solution Approach 1:
By segmenting the arbitration function across multiple distributed nodes, each handling a local subset of sources, the system achieves better bandwidth utilization and fairness without requiring any single node to manage all sources, thus controlling overall complexity.
Solution Approach 2:
Each distributed arbiter is optimized for its local context, managing arbitration for nearby sources with tailored logic. This local optimization improves overall system productivity and fairness while keeping individual component complexity low, as each arbiter only needs to handle its local segment rather than the entire system.
3Device complexity
If data packets are buffered at a centralized controller, then arbitration control is simplified, but transmission latency increases due to the additional buffering step and centralized processing
Solution Approach 1:
The patent distributes buffering and control functions across multiple arbiters along the pipeline rather than concentrating them at a single point. This segmentation allows data packets to be buffered and processed closer to their sources, increasing transmission speed while each arbiter maintains manageable control logic for its local segment.
Solution Approach 2:
Each distributed arbiter acts as an intermediary between local data sources and the shared resource, performing local buffering and arbitration. This eliminates the need for long-distance buffering at a centralized controller, improving transmission speed while keeping control logic distributed and manageable at each intermediary node.
4Reliability
If round-robin scheduling is used for arbitration, then fairness among data sources is improved, but latency increases due to the sequential nature of the scheduling algorithm
Solution Approach 1:
The patent segments the round-robin scheduling function across multiple distributed arbiters, each handling a local subset of sources. This allows parallel arbitration operations in different segments, maintaining fairness within each segment while reducing overall arbitration delay through concurrent processing.
Solution Approach 2:
The patent transforms the single-dimensional sequential round-robin scheduling into a multi-dimensional parallel structure by distributing arbitration across multiple spatial locations. Each segment performs its own round-robin scheduling independently, achieving fairness while reducing total arbitration delay through parallel execution across different dimensions of the system.
Data Source
AI summary
Passage of data packets on a data pipeline is arbitrated in a distributed manner along the pipeline. Multiple data arbiters each operate to merge data from a respective data source to the data pipeline at a distinct point in the pipeline. At each stage, a multiplexer selectively passes, to the data pipeline, an upstream data packet or a local data packet from the respective data source. A register stores an indication of data packets passed by the multiplexer based on the respective data source originating the data packet. A controller controls the multiplexer to select the upstream data packet or the local data packet based on the indication of data packets passed by the multiplexer.