Atomic Multicast via Remote Write Sequencing

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current multicasting technologies face high latency and resource inefficiencies due to the need for distributed agreement protocols and centralized management, which are not suitable for scalable and low-latency applications like radio access networks and cloud computing.

Innovation Solution

Implementing an atomic multicast system using remote write operations for direct memory communication between processing units, with a sequencer managing message ordering and acknowledgments to ensure all nodes receive messages in the same order, reducing overhead and latency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If distributed agreement protocols and centralized management are used for multicasting, then message ordering and reliability are improved, but latency increases and scalability deteriorates

Engineering Contradiction:
Improvemessage orderingVSAvoidlatency
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The system segments the multicast group into sender nodes and receiver nodes with distinct roles. Senders independently generate messages while receivers independently collect and order messages. This segmentation eliminates the need for centralized coordination of message ordering, reducing latency while maintaining reliability through local ordering at each receiver.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Each receiver node autonomously collects messages from multiple senders and independently manages its own message ordering using a queue data structure. The receiver self-services the ordering function without requiring centralized control or waiting for global agreement, thereby reducing latency while ensuring reliable ordered delivery to the application.

Inventive Principle:
Principle #25Self-service

2Reliability

If distributed agreement protocols are used for multicasting, then message ordering is improved, but device complexity and resource consumption increase

Engineering Contradiction:
Improvemessage orderingVSAvoidprotocol overhead
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent extracts the message ordering function from the centralized management layer and places it at the individual receiver level. Each receiver maintains its own queue and independently orders messages, eliminating complex inter-node coordination protocols. This extraction simplifies the overall system architecture while maintaining reliable ordered delivery.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

Instead of having centralized authority control message ordering, the patent inverts the approach by giving each receiver autonomous control over its own message ordering. This inversion eliminates the need for complex distributed agreement protocols and centralized management overhead, reducing device complexity while maintaining reliability.

Inventive Principle:
Principle #13The other way round (Inversion)

3Reliability

If centralized management is used for multicasting, then coordination is improved, but scalability deteriorates

Engineering Contradiction:
ImprovecoordinationVSAvoidscalability
Core Design Contradiction:
ReliabilityVSAdaptability or versatility

Solution Approach 1:

The system segments coordination responsibilities by eliminating centralized management entirely. Each receiver independently coordinates its own message collection and ordering operations. This segmentation allows the system to scale horizontally by adding more receivers without increasing coordination overhead, as each new receiver independently manages its own message processing.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Each receiver node performs self-coordination by autonomously collecting messages from senders and managing its own message ordering queue. This self-service approach eliminates the need for centralized coordination mechanisms, enabling the system to scale to large numbers of receivers while maintaining reliable coordinated message delivery through local autonomous operations.

Inventive Principle:
Principle #25Self-service

Data Source

PatentUS12081357B2Ultra-low latency atomic multicast
Publication Date: 2024.09.03 TELEFONAKTIEBOLAGET LM ERICSSON (PUBL)
  • US12081357B2 patent drawing
  • US12081357B2 patent drawing
  • US12081357B2 patent drawing

AI summary

A method is implemented by a computing device for atomic multicast support for a multicast group, includes receiving a message order from a sequencer via remote write, the message order defining a sequence for a plurality of messages from a multicast group, receiving the plurality of messages from at least one sender in the multicast group via remote write, and providing the plurality of messages to an application in the message order, in response to determining that all of the plurality of messages have been received.