Time Division Multiplexing Arbitration for Shared Optical Links
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
As semiconductor-based microprocessors approach physical limits, increased heat dissipation and power consumption prompt the need for alternative designs, such as parallel processing systems with multiple microprocessors, which require higher I/O bandwidth and efficient data transmission methods to manage independent tasks and shared information.
Innovation Solution
A method and system for arbitration in a macrochip architecture using a silicon photonic network that involves selecting a sending node, broadcasting arbitration requests, and incrementing counters to manage data transmission through optical data channels, enabling efficient data transfer between nodes within arbitration domains.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple microprocessors work in parallel to surpass physical limits of single processor, then processing capability is improved, but I/O bandwidth requirements increase significantly
Solution Approach 1:
The patent segments the I/O bandwidth into multiple shared optical links that are time-multiplexed among multiple processors. Instead of providing dedicated high-bandwidth links to each processor, the system divides the optical links into time slots that are allocated to different processors sequentially, thereby reducing the total I/O bandwidth requirement while maintaining parallel processing capability.
Solution Approach 2:
The system implements periodic time-division multiplexing where shared optical links are allocated to different processors in periodic time slots. Each processor receives dedicated access to the optical links during its allocated timeslot, creating a periodic pattern of resource allocation that reduces overall bandwidth requirements while maintaining effective parallel processing.
2Quantity of substance
If time division multiplexing is used to share optical links among multiple processors, then I/O bandwidth requirements are reduced, but arbitration complexity increases
Solution Approach 1:
The patent implements a self-service arbitration mechanism where each processor maintains local counters that automatically track its allocated timeslots and manage its own access to shared optical links. The arbitration logic is distributed across processors rather than centralized, with each processor independently determining when it should transmit based on its counter values, thereby reducing overall arbitration complexity.
Solution Approach 2:
The system performs preliminary allocation of time slots to each processor before actual data transmission occurs. Counters are pre-configured with timeslot information, and processors use these pre-computed values to automatically determine transmission opportunities without requiring complex real-time arbitration decisions during data transfer.
3Quantity of substance
If dedicated optical links are provided to each processor, then I/O bandwidth is sufficient, but system cost and complexity increase
Solution Approach 1:
The patent merges multiple dedicated optical link connections into shared optical links that serve multiple processors. By combining the functionality of what would otherwise require separate dedicated links for each processor into shared infrastructure with time-division multiplexing, the system reduces the total number of optical components and interconnections required, thereby lowering system cost and complexity.
Solution Approach 2:
The shared optical links are designed to serve multiple functions and multiple processors universally. Rather than having specialized dedicated links for each processor, the same optical infrastructure is universally shared across all processors through time-multiplexed access, reducing the overall system complexity and component count while maintaining sufficient I/O bandwidth capacity.
Data Source
AI summary
A method for arbitration including selecting, for an arbitration interval corresponding to a timeslot, a sending node from a plurality of sending nodes in an arbitration domain, where the plurality of sending nodes include a plurality of source counters; broadcasting, by the sending node and in response to selecting the sending node, a transmitter arbitration request for the timeslot during the arbitration interval; receiving, by the plurality of sending nodes, the transmitter arbitration request; incrementing the plurality of source counters in response to receiving the transmitter arbitration request; and sending, during the timeslot, a data item from the sending node to a receiving node via an optical data channel.