Parallel Process Group Identifier Generation Without Communication

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

In parallel computing, managing group identifiers for processes across multiple compute nodes is inefficient due to the need for communication among processes to establish operational groups, which increases overhead and complexity.

Innovation Solution

Each process generates a unique group identifier independently based on a list of process identifiers without communication, using algorithms such as stride triplets or hashing, to eliminate the need for central management and reduce traffic.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If processes communicate to establish operational groups and exchange group identifiers, then group establishment is achieved, but communication overhead and system complexity increase

Engineering Contradiction:

Improvegroup establishmentVSAvoidcommunication overhead

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

Each process independently generates its own group identifier based on local information (process identifier and group parameters) without requiring communication with other processes. This self-service approach eliminates the need for centralized group identifier management and peer-to-peer communication protocols, directly resolving the technical contradiction by achieving reliable group establishment through independent local computation

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The system pre-establishes deterministic relationships between process identifiers, group parameters, and group identifiers through mathematical algorithms (hashing, stride triplets). This preliminary action creates a lookup table or formula that allows any process to immediately determine its group identifier without real-time communication, eliminating communication overhead while ensuring consistent group assignment across all processes

Inventive Principle:

Principle #10Preliminary action

2Stability of the object's composition

If centralized management of group identifiers is implemented, then consistency is maintained, but communication traffic and management complexity increase

Engineering Contradiction:

Improvegroup identifier consistencyVSAvoidcommunication traffic

Core Design Contradiction:

Stability of the object's compositionVSQuantity of substance

Solution Approach 1:

The invention transforms the management approach from centralized storage and distribution of group identifiers to a parameter-based deterministic generation system. By changing the parameters (process identifier, number of groups, stride) and applying a consistent algorithm, all processes generate identical group identifiers without communication, maintaining consistency while eliminating communication traffic

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent introduces a mathematical algorithm (hashing function or stride triplet calculation) as an intermediary between process identifiers and group identifiers. This intermediary deterministically maps process parameters to group identifiers, replacing the need for centralized management communication while ensuring consistent results across all processes in the parallel system

Inventive Principle:

Principle #24Intermediary (Mediator)

3Productivity

If processes generate group identifiers independently without communication, then communication overhead is reduced, but challenges in ensuring unique and consistent identifiers arise

Engineering Contradiction:

Improvegroup establishment efficiencyVSAvoididentifier uniqueness

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The invention uses copying in the form of deterministic algorithm replication. Each process copies the same group identifier generation algorithm and inputs (process identifier, group parameters) to produce identical outputs. This copying approach ensures uniqueness and consistency without communication, as the deterministic nature of the algorithm guarantees that the same inputs always produce the same unique group identifier

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The patent replaces the mechanical communication system (message passing, centralized management) with a mathematical computation system. By substituting communication mechanics with deterministic mathematical algorithms (hashing, arithmetic operations on stride triplets), the system achieves identifier generation with guaranteed uniqueness and consistency without requiring inter-process communication infrastructure

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS9473545B2Administering group identifiers of processes in a parallel computer

Publication Date: 2016.10.18 LENOVO ENTERPRISE SOLUTIONS (SINGAPORE) PTE LTD

AI summary

Administering group identifiers of processes in a parallel computer includes each process in a set of processes, receiving from a compute node of the plurality of compute nodes, a request to establish the set of processes as an operational group including receiving a list of process identifiers for each process of the set of processes. Embodiments also include each process generating without communication amongst the processes, a unique group identifier in dependence upon the list of process identifiers.