What is Juice?
Juice offers GPU-over-IP technology. With a quick software installation, your GPUs transform into remote virtual resources. These resources can be pooled, shared, and accessed by AI and graphics workloads.
The software enables a new computing substrate where accelerated compute flows over networks. Applications can tap into GPUs from anywhere over existing standard networks without changes to applications or hardware. This allows to share GPUs with multiple application hosts remotely and to allocate them dynamically.
Features
- Remote GPU Access: Access GPUs from anywhere over standard networks.
- Seamless GPU Sharing: Share GPUs with multiple application hosts remotely and allocate dynamically.
- Dynamic Resource Allocation: Adapt and re-balance compute resources dynamically.
- Orchestration without k8s: Dynamic GPU scheduling, pooling, and fractionation without Kubernetes.
- Cross-Cloud Compatibility: Route GPU capacity from anywhere it is more abundant to anywhere it is scarce.
- GPU flexibility: Dynamic sharing and pooling software gives any GPU-accelerated workload direct remote access to any GPU capacity on any network.
Use Cases
- Dynamic GPU scheduling in superclusters.
- Routing GPU capacity across different cloud environments.
- Pooling and sharing GPU capacity across teams or departments.
- Enabling data scientists to run AI workloads dynamically across distributed GPU networks.
