Remote Write Communication Channel for Low Latency Inter-Process Data Transfer

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current inter-process communication methods, particularly in large-scale computing systems, face significant latency and overhead due to the reliance on network protocols, which hinder efficient data sharing between processing units and are not scalable for latency-sensitive applications.

Innovation Solution

Implementing a low-latency communication channel using remote writes that allow direct memory access between processing units, leveraging technologies like PCIe and RDMA to bypass network protocol handlers and reduce latency by writing data directly to receive buffers.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If standard network protocols are used for communication between processing units, then flexibility and interconnect bandwidth are provided, but latency and performance overhead increase significantly

Engineering Contradiction:

Improvecommunication flexibilityVSAvoidcommunication latency

Core Design Contradiction:

Adaptability or versatilityVSLoss of time

Solution Approach 1:

The patent extracts the network protocol handling functionality from the communication path by implementing remote direct memory access (RDMA) operations. This allows data to be transferred between processing units without being processed through network protocol handlers, effectively removing the protocol overhead layer while maintaining network connectivity capabilities.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent introduces a communication channel with sequence number tracking as an intermediary mechanism between sending and receiving applications. This intermediary manages the communication state locally at each endpoint, enabling efficient data transfer while maintaining the necessary coordination without requiring full network protocol processing.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Adaptability or versatility

If network protocol handlers are implemented on each processing unit, then communication flexibility is maintained, but compute resources are significantly consumed

Engineering Contradiction:

Improveprotocol handling capabilityVSAvoidcompute resource consumption

Core Design Contradiction:

Adaptability or versatilityVSUse of energy by moving object

Solution Approach 1:

The patent removes network protocol handlers from the processing units by implementing remote write operations that bypass protocol processing entirely. The communication channel manages necessary protocol functions through local state tracking and sequence numbers, eliminating the need for dedicated protocol handler software on each processing unit.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The communication channel implements self-service mechanisms through local tracking of sequence numbers and buffer positions at each endpoint. Each processing unit independently manages its own communication state without requiring external protocol processing, reducing compute resource consumption while maintaining communication functionality.

Inventive Principle:

Principle #25Self-service

3Speed

If shared memory is used for inter-process communication within a processing unit, then communication speed is fast, but shared memory is only available on a single processing unit

Engineering Contradiction:

Improvecommunication speedVSAvoidcross-processing-unit capability

Core Design Contradiction:

SpeedVSAdaptability or versatility

Solution Approach 1:

The patent creates a universal communication mechanism that functions like shared memory within a processing unit but extends across multiple processing units. The remote write channel provides shared memory-like semantics (direct address access, sequential writes) while enabling communication between any processing units in the system, combining the speed of shared memory with the versatility of network communication.

Inventive Principle:

Principle #6Universality (Multi-functionality)

4Adaptability or versatility

If conventional network protocols are used for data sharing between processing units, then communication is possible, but latency is a magnitude higher than shared memory communication

Engineering Contradiction:

Improvedata sharing capabilityVSAvoidlatency compared to shared memory

Core Design Contradiction:

Adaptability or versatilityVSLoss of time

Solution Approach 1:

The patent substitutes the mechanical network protocol processing system with a more efficient memory access mechanism. By using remote direct memory access operations with local state tracking, the system replaces the multi-layer protocol processing mechanism with a streamlined path that achieves latency closer to shared memory communication while maintaining the ability to share data across processing units.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS11641288B2Low latency small message communication channel using remote write

Publication Date: 2023.05.02 TELEFONAKTIEBOLAGET LM ERICSSON (PUBL)

AI summary

A method implemented by an electronic device for sending data on low latency communication includes remote writing a sequence number of a message to be sent next to a receiver, determining whether there is an open position in a receive buffer of the receiver using a local tracking mechanism, writing data of the message to an area of an address space of a sending application that is mapped onto the receive buffer as a result of determining there is the open position in the receive buffer, incrementing a local sequence counter, and updating position information in the local tracking mechanism.