Vultr, the world’s largest privately held cloud computing platform, is unveiling the Vultr GPU Stack and Container Registry, tackling the issues associated with configuring and provisioning GPUs, as well as the daunting task of ensuring that AI and machine learning (ML) models are secure and compliant.
The Vultr GPU Stack supports instant provisioning of the full scope of NVIDIA GPUs, pre-configured with the NVIDIA CUDA Toolkit, NVIDIA cuDNN, and NVIDIA drivers. These GPUs are instantly deployable, alleviating the pains of GPU configuration which often necessitate calibration to specific model requirements for each application.
“For those trained in ‘the dark art of configuring GPUs,’ configuring and provisioning them is still a notoriously painful activity,” said Kevin Cochrane, CMO at Vultr. “Vultr knew firsthand that this was an acute problem and launched the GPU Stack to provide relief and eliminate resulting bottlenecks.”
Vultr is also debuting its Kubernetes-based Container Registry, engineered to give organizations access to NVIDIA ML models from the NVIDIA NGC catalog across Vultr’s 32-plus cloud data center locations, across all six continents.
The private registry combines public models with an enterprise’s private datasets, allowing developers to train models based on relevant proprietary data, further creating their own instance of the model for inference. That model is then accessible to authorized users via each company’s individual private container registry.
The result is fast global instantiation and AI model tuning, all synchronized throughout private registries in each region.
“Now, every data science and engineering team can get cloud GPU access on their own terms, anywhere they are based in the world. But Vultr did not stop at access—they are pushing innovation much further with the launch of their Container Registry,” said Cochrane. “The Registry is essentially a customizable model catalog—a private model garden, for globally distributed teams to build on each others' model work, and train and scale their models on their own data sets.”
Regarding security, the “Vultr Container Registry secures artifacts with policies and role-based access control, and ensures images are scanned and free from vulnerabilities, and signs images as trusted,” said Cochrane.
“Based on Harbor, a CNCF Graduated project, Vultr Container Registry delivers compliance, performance, and interoperability to help you consistently and securely manage artifacts across cloud-native compute platforms like Kubernetes and Docker,” he continued.
To learn more about Vultr’s GPU Stack and Container Registry, please visit https://www.vultr.com/.