HELIOS Optical Switch Distributed Scheduling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current switching systems, particularly optical switches, face challenges in achieving 100% throughput and efficient power consumption due to limitations in packet buffering and scheduling algorithms, especially under high traffic demands and varying traffic patterns.
Innovation Solution
The High Energy-efficiency Locally-scheduled Input-queued Optical Switch (HELIOS) employs a distributed scheduling method that uses local queue information and minimal message passing between input and output ports, generating a Hamiltonian walk schedule to optimize crosspoint scheduling and reduce power consumption, achieving 100% throughput under any admissible Bernoulli traffic matrix.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If electronic switching systems are used to support high traffic demands, then bandwidth and switching capacity are improved, but power consumption increases significantly
Solution Approach 1:
The system is segmented into electronic control planes and optical data planes. The electronic portion handles only scheduling decisions and control signaling, while the optical fabric handles high-speed packet switching without conversion, dividing functions to minimize electronic power consumption while maintaining high throughput capacity.
Solution Approach 2:
The patent replaces electronic packet switching with optical packet switching in the fabric. By using optical signals instead of electrical signals for packet transmission and switching, the system achieves high bandwidth with significantly reduced power consumption, as optical components consume less power than electronic components for high-speed data handling.
2Use of energy by moving object
If optical switching fabric is used to reduce power consumption, then energy efficiency is improved, but packet buffering capability deteriorates due to lack of optical buffers
Solution Approach 1:
The patent introduces optical buffers as intermediary components in the optical fabric. These buffers enable temporary storage of optical packets without conversion to electrical domain, maintaining the low power consumption advantage of optical switching while providing necessary buffering capability for traffic regulation and congestion management.
Solution Approach 2:
The optical fabric is designed to perform multiple functions including packet switching, packet buffering, and traffic management all in the optical domain. This multi-functionality eliminates the need for separate electronic buffering stages, maintaining energy efficiency while providing comprehensive packet handling capabilities.
3Productivity
If centralized scheduling algorithms are used to achieve 100% throughput, then throughput is improved, but system complexity and message passing requirements increase
Solution Approach 1:
The centralized scheduler is segmented and distributed to line cards. Each line card executes local scheduling algorithms based on distributed information, eliminating the need for a complex centralized scheduling system while achieving comparable throughput performance through parallel distributed decision-making.
Solution Approach 2:
Line cards are empowered with self-service scheduling capabilities. Each line card independently makes scheduling decisions using local queue information and distributed coordination protocols, reducing system complexity by eliminating centralized control while maintaining throughput through autonomous local optimization.
4Device complexity
If distributed scheduling is used to reduce complexity, then device complexity is reduced, but throughput may not reach 100% under varying traffic patterns
Solution Approach 1:
The distributed scheduling system incorporates feedback mechanisms where line cards exchange queue status information and scheduling decisions with neighboring line cards. This feedback enables coordinated distributed scheduling that adapts to varying traffic patterns, achieving 100% throughput through iterative optimization without centralized control.
Solution Approach 2:
The distributed scheduling algorithm is designed to be dynamic and adaptive to changing traffic conditions. Line cards continuously adjust scheduling decisions based on real-time queue states and traffic patterns, enabling the system to maintain high throughput under varying loads without the rigidity of centralized control.
Data Source
AI summary
Scheduling methods and apparatus for use with optical switches with hybrid architectures are provided. An exemplary distributed scheduling process achieves 100% throughput for any admissible Bernoulli arrival traffic. The exemplary distributed scheduling process may be easily adapted to work for any finite round trip time, without sacrificing any throughput. Simulation results also showed that this distributed scheduling process can provide very good delay performance for different traffic patterns and for different round trip times associated with current switches.