Multibus Optical Interconnect Flow Control

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Data center switches face challenges in increasing bandwidth without escalating power consumption and system cost, particularly due to signal integrity limitations and the need to prevent buffer overflow in multibus systems.

Innovation Solution

Implementing multibus optical interconnect fabrics with periodic distribution of buffer status information to prevent buffer overflow, using optical signals and flow control methods that allow each node to monitor and manage buffer status across the system, thereby preventing data loss or corruption.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If bandwidth is increased to meet data center demands, then data transmission capacity is improved, but power consumption increases and signal integrity deteriorates

Engineering Contradiction:
ImprovebandwidthVSAvoidinterconnect power consumption
Core Design Contradiction:
ProductivityVSUse of energy by moving object

Solution Approach 1:

The patent replaces electronic interconnects with optical interconnects in the multibus system. Optical signals are used to transmit data across the fabric, substituting the mechanical/electronic signal transmission medium. This enables higher bandwidth capacity while reducing power consumption compared to traditional electronic interconnects, as optical transmission is more energy-efficient for high-speed data transfer.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Productivity

If bandwidth is increased to meet data center demands, then data transmission capacity is improved, but system cost increases

Engineering Contradiction:
ImprovebandwidthVSAvoidsystem cost
Core Design Contradiction:
ProductivityVSEase of manufacture

Solution Approach 1:

The patent implements a universal flow control mechanism that operates across all links in the multibus system. The same buffer status information distribution protocol is used throughout the entire fabric, allowing a single implementation to manage flow control for all data paths. This universal approach avoids the need for separate flow control mechanisms on each link, reducing overall system complexity and cost while maintaining high bandwidth capability.

Inventive Principle:
Principle #6Universality (Multi-functionality)

3Reliability

If point-to-point flow control is used on each link, then data loss is prevented, but buffer requirements increase and system complexity grows

Engineering Contradiction:
Improvedata loss preventionVSAvoidflow control buffer requirements
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent merges the flow control function into a centralized fabric-wide mechanism rather than implementing separate point-to-point flow control on each link. Buffer status information is distributed across the entire multibus fabric, and nodes make flow control decisions based on global buffer status rather than local link conditions. This consolidation reduces the buffer requirements on individual links while maintaining data loss prevention through coordinated fabric-wide flow control.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentUS8391717B2Flow-control methods and systems for multibus systems
Publication Date: 2013.03.05 VALTRUS INNOVATIONS LTD
  • US8391717B2 patent drawing
  • US8391717B2 patent drawing
  • US8391717B2 patent drawing

AI summary

Methods and systems are provided that prevent buffer overflow in multibus systems. In one aspect, a method for controlling the flow of data in a multibus system includes, for each node having an associated broadcast bus in the multibus system, generating status information regarding available data storage space of each receive buffer of the node. The method includes broadcasting the status information to the other nodes connected to the broadcast bus and collecting status information regarding the available storage space of receive buffers of the other nodes connected to the broadcast bus. The method also includes determining whether or not to send data from the node to at least one of the other nodes over the broadcast bus based on the collected status information.