Azure Storage innovations: Unlocking the future of data

Whether you are improving the resilience of mission-critical workloads oregon modernizing bequest systems; Azure Storage has a solution for you.

Microsoft is redefining what’s imaginable successful the nationalist unreality and driving the adjacent question of AI-powered translation for organizations. Whether you’re pushing the boundaries with AI, improving the resilience of mission-critical workloads, oregon modernizing bequest systems with cloud-native solutions, Azure Storage has a solution for you.

At Microsoft Ignite 2025 and KubeCon North America past month, we showcased the latest innovations successful Azure Storage, powering your workloads. Here is simply a recap of those releases and advancements.

Innovating for the aboriginal with AI

Azure Blob Storage provides a unified retention instauration for the full AI lifecycle, powering everything from ingestion and mentation to checkpoint absorption and exemplary deployment.

To alteration customers to rapidly train, fine-tune, and deploy AI models, we evolved the Azure Blob Storage architecture to standard and present exabytes of capacity, 10s of Tbps throughput, and millions of IOPS to GPUs. In this video, you tin spot a azygous retention relationship scaling to implicit 50 Tbps on work throughput. Azure Blob Storage is besides the instauration that enables OpenAI to bid and service models astatine unprecedented velocity and scale. 

Fig 1. Storage-centric presumption of AI grooming and fine-tuning

For customers handling terabyte oregon petabyte standard AI grooming data, Azure Managed Lustre (AMLFS) is simply a high-performance parallel record strategy delivering monolithic throughput and parallel I/O to support GPUs continuously fed with data. AMLFS 20 (preview) supports 25 PiB namespaces and up to 512 GBps throughput. Hierarchical Storage Management (HSM) integration enhances AMLFS scalability by enabling seamless information question betwixt AMLFS and your exabyte-scale datasets successful Azure Blob Storage. Auto-import (preview) allows you to propulsion lone required datasets into AMLFS, and auto-export sends trained models to semipermanent retention oregon inferencing.

Rakuten is accelerating the grooming of Japanese ample connection models connected Microsoft Azure, leveraging Azure Managed Lustre, Azure Blob Storage, and Azure Kubernetes Service to maximize GPU utilization and simplify scaling.

Natalie Mao, VP, AI & Data Division, Rakuten Group

Once models are trained and fine-tuned, inferencing takes halfway signifier delivering real-time predictions and insights. Azure Blob Storage provides best-in-class retention for Microsoft AI services, including Microsoft Foundry Agent Knowledge (preview) and AI Search retrieval agents (preview), enabling customers to bring their ain retention accounts for afloat flexibility and control, ensuring that endeavor information remains unafraid and acceptable for retrieval-augmented procreation (RAG).

Additionally, Premium Blob Storage delivers accordant low-latency and up to 3X faster retrieval performance, captious for RAG agents. For customers that similar open-source AI frameworks, Azure Storage built LangChain Azure Blob Loader which delivers granular security, memory-efficient loading of millions of objects and up to 5x faster show compared to anterior assemblage implementations.

Fig 2. Storage-centric presumption of AI inference with endeavor data

Azure Storage is evolving to beryllium an integrated, intelligent AI-driven level simplifying absorption of exabyte-scale AI data. Storage Discovery and Copilot enactment unneurotic to assistance you analyse and recognize however your information property is evolving implicit clip utilizing dashboards and questions successful earthy language. With Storage Discovery and Storage Actions, you tin optimize costs, support your information and govern ample datasets with hundreds of billions of objects utilized for training, and fine-tuning.

Optimizing modern applications with Cloud Native

Modern cloud-native applications request agility. Two principles consistently basal out: elasticity and flexibility. Your retention should standard seamlessly with dynamic workloads—without operational overhead. The innovations beneath are designed for the cloud, enabling you to auto-scale, optimize costs intelligently, and present the show needed by modern applications.

Azure Elastic SAN provides cloud-native artifact retention for standard and choky Kubernetes integration for accelerated scaling, and multi‑tenancy that optimizes cost. With caller auto scaling support, Elastic SAN automatically expands resources arsenic needed, making it easier to negociate retention footprints crossed workloads. Early adjacent year, we’ll widen Kubernetes integration via Azure Container Storage for Azure Kubernetes Service (AKS) to wide availability (GA). These enhancements fto you support acquainted hosting environments portion layering successful cloud-native capabilities.

Cloud-native agility is besides captious for modern applications built connected entity storage, with the request to optimize costs and show for dynamic and unpredictable postulation patterns. Smart Tier (preview) connected Azure Blob Storage continuously analyzes entree patterns, moving information betwixt tiers automatically.

New information starts successful the blistery tier. After 30 days of inactivity, it moves to cool, and aft 90 days, to cold. If an entity is accessed again, it’s promoted backmost to blistery which keeps information successful the astir cost-effective tier automatically. You tin optimize costs without sacrificing performance, simplifying information absorption astatine standard and keeping your absorption connected building.

Hosting mission-critical workloads

Enterprises contiguous tally mission-critical workloads that necessitate artifact retention and present predictable show and uncompromising concern continuity. Azure Ultra Disk is our highest-performance artifact retention offering, purpose-built for workloads similar high-frequency trading, ecommerce platforms, transactional databases, and physics wellness grounds systems that request exceptional speed, reliability, and scalability.

With Azure Ultra Disk, we tin confidently standard our level globally, knowing that show and resilience volition conscionable endeavor expectations, that consistency allows our teams to absorption connected AI innovation and workflow automation alternatively than infrastructure.

Charles McDaniels, Director of Systems Engineering Management for Global Cloud Services, ServiceNow

We cognize performance, cost, and concern continuity stay the apical priorities for our customers and we are raising the barroom successful each category:

• Performance: We person further improved the mean latency for Azure Ultra Disk by 30% with mean latency good nether 0.5ms for tiny IOs connected virtual machines (VMs) with Azure Boost. A azygous Azure Ultra Disk tin present manufacture starring show of 400K IOPS and 10 GBps throughput. In addition, with Ebsv6 VMs, some Premium SSD v2 and Azure Ultra Disk tin present manufacture starring VM show standard of 800K IOPS and 14 GBps throughput for the astir demanding applications.
• Cost: Flexible provisioning for Azure Ultra Disk reduces full outgo of ownership by up to 50%, letting you standard capacity, IOPS, and MBps independently astatine finer granularity.
• Business continuity: Instant Access Snapshots (preview) lets you backup and reconstruct your workloads instantly with exceptional show connected rehydration. This differentiated acquisition for Azure Premium v2 and Ultra Disk helps destruct the operational overhead of monitoring snapshot readiness oregon pre‑warming resources, portion reducing recovery, refresh, and scale‑out times from hours to seconds.

Azure NetApp Files (ANF) is designed to present debased latency, precocious performance, and information absorption astatine scale. Its ample volumes capabilities person been importantly expanded providing an implicit 3x increase successful azygous measurement capableness standard to 7.2 PiB and a 4x increase successful throughput to 50 GiBps. Cache volumes bring information and files person to wherever users request accelerated entree successful a abstraction businesslike footprint. These marque ANF suitable for respective high-performance computing workloads specified arsenic Electronic Design Automation (EDA), Seismic Interpretation and Visualization, Reservoir Simulations, and Risk Modeling. Microsoft is not lone positioning ANF for ngo captious applications but besides utilizing ANF for in-house silicon design.

Breaking barriers—migrating your retention infrastructure

Every organization’s unreality travel is unique. Whether you request to determination existing environments to the unreality with minimal disruption oregon program a afloat modernization, Azure Storage offers solutions for you. Storage Migration Solution Advisor successful Copilot tin supply recommendations to assistance streamline the decision-making process for these migrations. 

Azure Data Box and Storage Mover simplify the migration travel from on-premises and different clouds to Azure. The adjacent procreation Azure Data Box is present GA. Storage Mover is our afloat managed information migration work that is secure, businesslike and scalable with caller capabilities: on-premises NFS shares to Azure Files NFS 4.1, on-premises SMB shares to Azure Blob storage, and cloud-to-cloud transfers. Storage Migration Solution Advisor successful Copilot accelerates decision-making for migrations.

For users acceptable to migrate their NAS information estates, Azure Files now makes this easier than ever. We have introduced a new management model making it easier and more cost effectual to usage file shares. Additionally, Azure Files present enables you to eliminate complex on-premises Active Directory or domain controller infrastructure, with Entra-only identities for SMB shares. With unreality native identity support, you tin now manage your user permissions straight in Azure, including outer identities for applications like Azure Virtual Desktop (AVD).

Entra-only identities enactment with Azure Files transforms SLB’s Petrel workflows by removing dependencies connected on-premises domain controllers, simplifying individuality absorption and retention infrastructure for globally distributed teams moving connected analyzable exploration and reservoir characterization. This cloud-native architecture allows customers to entree SMB shares successful an casual and unafraid mode without analyzable VPN oregon hybrid infrastructure setups.

Swapnil Daga, Storage Architect for Tenant Infrastructure, SLB

ANF Migration Assistant simplifies moving ONTAP workloads from on-premises oregon different clouds to Azure. Behind the scenes, the Migration Assistant uses NetApp’s SnapMirror replication technology, providing efficient, afloat fidelity, block-level incremental transfers. You tin present leverage ample datasets without impacting accumulation workloads.

For customers moving on-premises spouse solutions who privation to migrate to Azure utilizing the aforesaid partner-provided technology, Azure has precocious introduced Azure Native offers with Pure Storage and Dell PowerScale.

To marque migrations easier, Azure Storage’s Migration Program connects you with a robust ecosystem of experts and tools. Trusted partners similar Atempo, Cirata, Cirrus Data, and Komprise tin accelerate migration of SAN and NAS workloads. This programme offers secure, low-risk transfers of files, objects, and artifact retention to assistance enterprises unlock the afloat imaginable of Azure.

Start your adjacent section with Azure Storage

The epoch of AI-powered translation is here. Begin your travel by exploring Azure’s precocious retention offerings and migration tools, designed to accelerate AI adoption, unreality migration, and modernization. Take the adjacent measurement contiguous and unlock caller possibilities with Azure Storage arsenic the instauration for your AI initiatives.

For immoderate questions, scope retired astatine azurestoragefeedback@microsoft.com.