openFuyao | 开源社区

Version Release

Check the Latest Version Release

Document Map

Knowledge for everyone from beginners to experts

Contribution Guide

Get Started with Contributing Quickly

Latest Events

Learn About the Latest Events

openFuyao: A Diversified Computing Cluster Software Ecosystem

openFuyao is an open-source community focusing on software technology innovation for general-purpose computing and intelligent computing clusters. We are dedicated to providing everyone with a diversified computing cluster software ecosystem to unleash the infinite value of intelligence. We aim to jointly create an efficient and innovative computing cluster community in which we bring together partners and developers through open collaboration and build diversified computing power interconnection and scheduling capabilities, unlock the ultimate computing power of clusters while ensuring security and reliability, develop key software root technologies for clusters, and establish unified and open technical standards.

Applications

Runtime component

Scheduling component

Compute enablement component

Container platform release

AI inference suite

Big data suite

Scenario acceleration suite

AI inference acceleration

Big data acceleration

Container platform

Authentication and authorization

Alarms and monitoring

Software repository management

Extension management

Multi-cluster management

Resource management

Application lifecycle management

Log management

Console

Compute enablement innovation

NUMA-aware scheduling

Ultra-large cluster scheduling

Colocation scheduling

Distributed job scheduling

Diverse compute enablement

KAE operator

NPU operator

Resource pool management

NPU software-based allocation

Node feature discovery

Installation and deployment

Container orchestration engine (Kubernetes)

Container runtime

Network runtime

Storage runtime

OSopenEuler | SUSE | Ubuntu | ...

Diverse computex86 | Arm | CPU | NPU | GPU | ...

openFuyao: Unlocking the potential of cluster resources and performance

Container Orchestration Engine Optimization

Low-noise container foundation is sued. The memory usage of Kubernetes components decreases by 30% in single-node deployment, supporting over 1,000 pods.

NUMA-aware Scheduling

NUMA intelligent scheduling is implemented for optimal resource allocation, improving payment application performance by 20%.

Resource Utilization Improvement

Colocation and scheduling of online and offline services improve cluster CPU usage by 30% and memory usage by 10%.

Distributed Job Scheduling

AI-powered distributed job scheduling improves TPS by 15% to 20% in typical inference scenarios.

AI Inference Acceleration

KVCache is supported and intelligent routing and cache hit strategy are optimized, reducing TTFT and breaking through GPU memory bottlenecks across N/S and E/W directions.

Ultra-Large Cluster Scheduling

High-performance scheduling is provided for clusters with more than 10,000 nodes, enabling task-based batch creation of pods.

Heterogeneous Hardware Resource Pooling

Node feature awareness is enabled through the integration of the KAE Operator and NPU Operator, enabling xPU resources to become available within seconds with efficient allocation.

Get started with openFuyao!