Interactive Access to Headless Kubernetes-Slurm Clusters

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Configuring and managing hybrid cluster deployments combining Kubernetes and Slurm for high-performance computing and Big Data jobs is a difficult and time-consuming process, as users typically require customized configurations that leverage features from both systems, which are not easily integrated.

Innovation Solution

An automated method is provided to create hybrid cluster deployments by combining Kubernetes with Slurm, utilizing a secure connection to enable interactive access to headless cluster managers, allowing seamless interaction between frameworks with and without software-defined networking.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If Kubernetes and Slurm are combined to create hybrid cluster deployments, then functionality and versatility are improved, but configuration complexity and time consumption increase

Engineering Contradiction:
ImprovefunctionalityVSAvoidconfiguration complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent introduces an intermediary component that bridges Kubernetes and Slurm frameworks, enabling them to work together without direct complex integration. This mediator handles the coordination and communication between the two systems, simplifying the configuration process while maintaining enhanced functionality.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The hybrid cluster deployment system is designed to provide universal functionality by combining the strengths of both Kubernetes (container orchestration, software-defined networking) and Slurm (job scheduling, HPC workloads) into a single multi-functional platform that can handle diverse computational tasks.

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

2Adaptability or versatility

If Kubernetes and Slurm are combined to create hybrid cluster deployments, then functionality and versatility are improved, but the configuration process becomes time-consuming

Engineering Contradiction:
ImprovefunctionalityVSAvoidconfiguration time
Core Design Contradiction:
Adaptability or versatilityVSLoss of time

Solution Approach 1:

The system performs preliminary configuration actions by pre-establishing integration templates, default settings, and automated provisioning scripts for hybrid cluster deployments. This allows the complex Kubernetes-Slurm integration to be configured quickly by leveraging pre-prepared configurations rather than building from scratch.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The hybrid cluster deployment system incorporates self-service capabilities through automated configuration tools that can autonomously set up and optimize the integration between Kubernetes and Slurm based on detected requirements and best practices, reducing manual configuration time.

Inventive Principle:
Principle #25Self-service

3Ease of operation

If a secure connection is created from Kubernetes to Slurm node, then interactive access is enabled, but network complexity increases

Engineering Contradiction:
Improveinteractive accessVSAvoidnetwork complexity
Core Design Contradiction:
Ease of operationVSDevice complexity

Solution Approach 1:

A network intermediary or gateway is introduced to manage the secure connection between Kubernetes and Slurm. This intermediary handles authentication, encryption, and connection management, enabling interactive access while abstracting the network complexity from users and applications.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentUS20250294014A1Interactive access to headless cluster managers
Publication Date: 2025.09.18 ADVANCED MICRO DEVICES INC
  • US20250294014A1 patent drawing
  • US20250294014A1 patent drawing
  • US20250294014A1 patent drawing

AI summary

A system and method for creating hybrid cluster deployments are disclosed. The system includes a first framework having a software networking layer, the first framework including a container network interface for accessing a client run on a pod of the first framework. A second framework has a head node and one or more working nodes configured in a cluster. A secure connection between the client in the first framework and the head node in the second framework. The secure connection configured to communicate requests from the client to the head node.