Visual Computing GPU Resource Pooling

Paraverse aims to create an interactive gateway to the digital parallel world. By integrating two cutting-edge fields—3D applications and decentralized infrastructure—along with underlying technologies such as real-time cloud rendering, cloud-native, and blockchain, Paraverse has developed several core technologies and capabilities. These aim to address the pain points in the production, distribution, and operation of 3D applications:

Visual Computing GPU Resource Pooling

The underlying architecture of Paraverse provides GPU resource pooling management capabilities by intercepting low-level graphics processing interfaces such as DX, OpenGL, and Vulkan. This allows the system to allocate seemingly exclusive GPU devices to applications and enable multiple graphics applications to share the GPU concurrently. By capturing post-rendered GPU image data from processes and extracting audio from audio devices for encoding, as well as isolating input/output devices and IO access at the process level for each application, Paraverse enables ultra-fine-grained GPU resource allocation—one machine can host multiple GPUs, and one GPU can be shared by multiple applications. The underlying platform supports both physical and virtual machines, and the operating system is compatible with Windows and Linux.