NVIDIA Unified Compute Framework (UCF) is a low-code framework for developing cloud-native, real-time, & multimodal AI applications. It features low-code design tools for microservices & applications, as well as a collection of optimized microservices and sample applications. Adopting a Microservices Architecture approach, Unified Compute Framework enables developers to combine microservices into cloud-native applications or services, meeting the real-time requirements of interactive AI use cases.
Each microservice has a bounded domain context (Vision AI, Conversational AI, Animation AI & Rendering, Data Analytics, etc.) and can be independently deployed, managed, & scaled within the application. The abstraction of each domain from the application reduces the need for low-level domain and platform knowledge. With UCF, developers can create complex AI applications in days rather than weeks and months. Moreover, application execution can be distributed on multiple devices and across the cloud to edge and embedded platforms.
Finally, a complete set of specifications and design guidelines will allow domain experts to create microservices using UCF-compatible NVIDIA SDKs.
Following are the main concepts in UCF, leveraging Kubernetes and Helm Charts constructs:
Microservice Specification standardizes how microservices are defined, including their endpoint specification, infrastructure requirements, and configurations, which helps improve microservices’ interoperability and allow quick & simple building of cloud-native applications by combining microservices published by different teams at NVIDIA as well as custom UCF microservices.
UCF Microservice is built on top of the application container image. In Kubernetes terms, it is generally a single Pod or Deployment with the main application container image and related other components.
UCF Application is created by connecting Microservices together. It can be represented in the form of a .yaml file with application-specific configuration for each microservice and compatible endpoint connections. Each application specifies the version of its microservices to improve overall application version control & dependency management.
Microservice Registry is a collection of UCF Microservices, bringing a rich set of features for low-code AI application development through UCF Studio. It supports NGC and local repositories to store microservices.
UCF tools help developers build and validate microservices and applications.
UCF Studio is a low-code NVIDIA Omniverse based IDE to browse microservices from the registry and build applications via a drag-and-drop interface. It also enables configuring as well as packaging of such apps to Helm Charts for deployment.
Microservice Builder CLI¶
UCF Microservice Builder CLI is a CLI tool to build and publish microservices.
The following diagram shows all the stages of the UCF workflow using the mentioned concepts and tools, mapped to different personas.