Google Cloud & VAST unite to advance hybrid AI for enterprise

Google Cloud and VAST Data have formed a partnership to offer enterprise artificial intelligence capabilities at scale across hybrid cloud environments. The arrangement enables organisations to use VAST Data's AI Operating System as a fully managed service on Google Cloud, supporting connections between cloud-based and on-premises infrastructure.

Hybrid data access

This collaboration enables businesses to maintain a unified environment for AI workloads, regardless of where their data or compute resources are located. Customers can now deploy clusters of servers across Google Cloud and private data centres without facing the extensive timelines and costs typically associated with data migrations.

The partnership allows for the creation of a single global namespace using VAST's AI Operating System and DataSpace technology. In a recent demonstration, VAST connected clusters separated by over 10,000 kilometres-linking infrastructure in the United States and Japan-to provide near real-time access to the same data while running AI inference tasks. This setup makes it possible to efficiently allocate workloads to resources, such as running inference on TPUs in one region and GPUs in another, without duplicating data or managing siloed environments.

Operational efficiency

The offering aims to address challenges businesses encounter when running AI workloads across multiple locations. Typically, fragmented storage environments and segmented data pipelines impede consistent, high-speed access to training data. Each environment change can also complicate compliance and governance.

"Through our partnership with Google Cloud, we're meeting customers where they are by delivering a fully managed AI OS. When leveraging our global namespace with intelligent streaming, Google Cloud customers can auto-deploy a VAST-managed cluster via Google Cloud Marketplace and start production in minutes, providing integrated governance and billing, elastic scale, and supporting the hybrid cloud mission - and it's all handled by VAST, making enterprise data instantly usable for agentic workloads," said Jeff Denworth, Co-Founder, VAST Data.

Performance integration

VAST's AI Operating System has demonstrated its ability to connect directly to Google Cloud's Tensor Processing Unit (TPU) virtual machines, facilitating large-scale AI workloads. During testing with Meta's Llama-3.1-8B-Instruct language model, the system achieved model load speeds similar to those of local NVMe disks, even during cold-start scenarios. The platform is engineered to maintain steady performance and support strict workflow demands, aiming to maximise the utilisation of AI accelerators.

"The VAST AI OS is redefining what it means to move fast in AI. Loading models at speeds comparable to local NVMe while delivering a full suite of advanced platform features," said Subramanian Kartik, Chief Scientist at VAST Data. "This is the kind of acceleration that turns idle accelerators into active intelligence, driving higher efficiency and faster time to insight for every AI workload."

Deployment options

Organisations can deploy production AI workloads on Google Cloud using their existing on-premises datasets without the typical need for migration planning or extended compliance cycles. The solution streams only required data subsets to AI models, reducing data replication and associated data movement costs, allowing organisations to focus resources on AI development and operations. Configuration also includes unified access controls and audit policies.

The global namespace provided by VAST's platform enables customers to migrate, replicate or cache data in Google Cloud, while maintaining consistent governance across all environments.

Unified platform

VAST's AI Operating System includes a suite of services, such as DataStore, DataBase, InsightEngine, AgentEngine, and DataSpace. These components scale across on-premises and Google Cloud environments, enabling teams to perform data preparation, training, inference, and analytics within a single framework. The system supports direct feeding of data to TPU virtual machines over validated NFS paths for model loads and offers small-file and metadata-optimised input/output.

VAST Data's deployment is available for Google Cloud customers, including those in Australia, where Google Cloud is estimated to hold around 20% of the market for cloud computing platforms. Joint validation and guidance are accessible for businesses and partners considering the deployment of a connected VAST DataSpace environment.

"Bringing VAST AI Operating System to Google Cloud Marketplace will help customers quickly deploy, manage, and grow the data solution on Google Cloud's trusted, global infrastructure," said Nirav Mehta, Vice President, Compute Platform at Google Cloud. "VAST can now securely scale and support customers on their digital transformation journeys."