Use Cases#
openYuanrong can serve as infrastructure for AI intelligent computing, used to develop AI inference, reinforcement learning, and other applications. It can also serve as infrastructure for general-purpose computing, developing big data analytics, HPC (High Performance Computing), and other applications.
openYuanrong as AI Infrastructure#
openYuanrong provides commonly used AI application frameworks through open and compatible approaches, simplifying end-to-end AI application development. Built-in heterogeneous distributed multi-level caching capabilities support efficient LLM training and inference. On-demand component-based flexible deployment facilitates integration with existing business.
Ecosystem Openness: Provides open application frameworks for different domains, and adapts commonly used frameworks like vLLM, Verl, etc. through adaptors, enabling zero-code modification access for existing applications.
Efficient Inference and Training: Built-in heterogeneous distributed multi-level caching capabilities allow inference to quickly transfer model parameters and KV Cache data, and training to switch between training and inference parameters with zero redundancy.
Flexible Deployment: Supports full or lightweight deployment, supports deployment on open-source Kubernetes or cloud clusters, facilitating integration with existing business systems.
AI Intelligent Computing Use Cases#
openYuanrong as General-Purpose Computing Infrastructure#
Microservices, big data analytics, and HPC are common distributed applications in general-purpose computing scenarios. openYuanrong’s multi-language function programming interfaces provide distributed parallelization capabilities for standalone programs, simplifying development complexity. Function-granularity application instances respond elastically to business at extreme speeds. Different application workloads support co-resource pool deployment, ensuring high performance while significantly improving resource utilization.
Simplified Development: openYuanrong’s multi-language function programming interfaces support commonly used development languages like Python, C++, and Java. By abstracting resources, adaptive dynamic scheduling hides complexities like elasticity and distributed scheduling, simplifying distributed application development.
Extreme Elasticity: Snapshot-based cold start acceleration capabilities allow function application instances to automatically scale at millisecond-level to respond to business traffic, eliminating the need to pre-allocate large resources for business peak demands, significantly improving resource utilization.
Diverse Workload Co-Pooling: openYuanrong clusters support co-resource pool deployment of distributed applications like microservices, big data analytics, and HPC, with efficient communication and data exchange, achieving high performance.