VAST Information has partnered with Microsoft to supply the VAST AI Working System (AI OS) natively on Microsoft Azure. This integration goals to create a high-performance, scalable AI infrastructure within the cloud, specializing in agentic AI and manufacturing mannequin workflows.
The VAST AI OS will run instantly on Azure’s infrastructure, permitting companies to make the most of acquainted Azure instruments, governance, safety, and billing. The purpose is to realize unified administration, constant efficiency, and Azure-level reliability whereas extending VAST’s on-premises options into Azure’s GPU-accelerated setting.
VAST Providers as a Native Azure Providing
Azure clients may have entry to VAST’s complete information companies stack, which incorporates unified storage, information cataloging, and database options. The platform is designed to help advanced AI processes throughout on-premises, hybrid, and multi-cloud environments with out requiring utility redesign.
By positioning VAST as a cloud-native Azure service, companies can consolidate AI information companies onto a single logical platform. This method goals to attenuate operational challenges and ship constant efficiency, information administration, and safety, no matter the place the information resides.
Jeff Denworth, VAST Information co-founder, described the partnership as a approach to create AI infrastructure that mixes efficiency, scalability, and ease of use for agentic AI. He emphasised that the Azure collaboration permits clients to unify information and AI processes with VAST’s acquainted instruments, now enhanced by Azure’s international presence and suppleness.
Agentic AI with InsightEngine and AgentEngine
The combination is optimized for agentic AI, by which autonomous brokers constantly function on dwell information. VAST AI OS on Azure will provide two key companies:
- VAST InsightEngine: This offers environment friendly, high-performance computing and database companies tailor-made for AI information duties, together with vector search, retrieval-augmented era (RAG), and information preparation. By working compute near information with improved I/O and indexing, it reduces latency and boosts throughput.
- VAST AgentEngine: This manages autonomous brokers on real-time information streams throughout hybrid and multi-cloud environments. This permits steady AI processing throughout globally distributed datasets with out the necessity for guide information staging or pipeline rewriting.
Collectively, these companies set up VAST AI OS on Azure as a management and execution layer for agentic AI workflows throughout on-premises and multi-cloud environments.
Efficiency at Scale for Mannequin Builders
VAST AI OS is designed to take care of busy GPU and CPU clusters throughout coaching and inference. On Azure, it would use the Laos VM Collection and Azure Enhance Accelerated Networking.
Key efficiency options embrace:
- Excessive-throughput information companies to totally make the most of GPU and CPU capabilities
- Clever caching for continuously accessed datasets to cut back lag time
- Metadata-optimized enter/output paths to deal with small-file and metadata-heavy AI duties
The purpose is to ship dependable efficiency from early exams by way of multi-region manufacturing rollouts, offering mannequin builders with a constant information layer as they develop.
Hybrid AI with Exabyte-Scale DataSpace
For purchasers with hybrid or on-premises VAST methods, the exabyte-scale VAST DataSpace provides a unified international namespace throughout websites and clouds. This single logical view removes information silos and eliminates the necessity for guide information copying.
Companies can broaden GPU-accelerated duties from on-premises to Azure with out typical information migration or reconfiguration. AI processes might be prolonged by focusing on the identical namespace, with VAST managing information placement and motion within the background.
Unified Information Entry and AI-Native Database
The Azure-integrated platform consists of:
- VAST DataStore: This helps file (NFS, SMB), object (S3), and block protocols from one platform. This setup permits legacy, analytics, and AI duties to share the identical information without having separate storage stacks.
- VAST DataBase: This combines quick transactional efficiency, warehouse-class question velocity, and information lake economics, enabling combined workloads to run on a single platform quite than separate OLTP, OLAP, and lake methods.
Elastic, Value-Environment friendly Structure
VAST’s Disaggregated, Shared-All the things (DASE) structure permits impartial scaling of computing and storage assets inside Azure. Prospects can regulate assets as fashions, datasets, and demand change, quite than over-allocating tied methods.
Constructed-in Similarity Discount additionally reduces storage necessities by eradicating duplicate information patterns, particularly for giant fashions, embeddings, and dataset variations. This leads to a less expensive basis for intensive AI infrastructure.
Alignment with Microsoft’s AI Roadmap
Aung Oo, Vice President of Azure Storage at Microsoft, acknowledged that VAST AI OS on Azure offers a high-performance, scalable platform constructed on the Laos VM Collection with Azure Enhance, extending on-premises AI pipelines into Azure’s GPU-supported infrastructure. He famous VAST’s present adoption by mannequin builders for scalability, efficiency, and AI-native options. He talked about that the partnership ought to simplify operations, decrease prices, and velocity up time-to-insight.
As Microsoft develops its AI infrastructure and customized silicon initiatives, VAST will work with the Azure workforce to satisfy future platform wants. This partnership positions VAST as a key information and AI platform, aiming to help future AI methods with an working system targeted on scalability, efficiency, and ease of use.
