On this article
Linux workloads are evolving rapidly as organizations modernize on-premises Linux functions, undertake cloud-native architectures, and construct new AI and data-intensive pipelines within the cloud. As this shift accelerates, groups want a managed file platform that delivers the efficiency, resilience, safety, and suppleness required to help these workloads in Azure.
Azure Information helps groups make that transition with totally managed file storage for Linux workloads. It combines acquainted file entry with built-in efficiency, information safety, safety, and Azure service integration, making it simpler to run shared file storage within the cloud with out managing file infrastructure instantly.
On this weblog, we discover how Azure Information NFS helps trendy Linux workloads throughout AI, cloud-native, enterprise modernization, and accomplice or ISV situations.
Velocity up AI and data-intensive workloads
AI inferencing is the place a educated mannequin is loaded and serves predictions. Earlier than any request is answered, the mannequin’s weights, which seize what the mannequin discovered throughout coaching, have to be learn from storage into the serving occasion. As fashions have grown to tens or lots of of gigabytes, that loading step closely influences how rapidly an endpoint begins and scales. It occurs each time a duplicate spins up, and since GPUs are among the many most costly assets within the stack, each minute they sit idle ready on weights is cash wasted.
Historically, groups embed mannequin weights instantly within the container picture or have every reproduction obtain its personal copy—producing bloated photos, duplicated information, and chilly begins that stretch to minutes simply when an inference fleet must scale quick. With Azure Information, the weights reside as soon as on a file share that each reproduction mounts and reads concurrently: photos keep lean, there’s a single copy to handle, and new replicas come on-line in seconds for small and mid-sized fashions. Help for the Linux NFS nconnect mount possibility permits a consumer to determine a number of parallel TCP connections to the identical file share, which might enhance throughput at scale.
The brand new Zonal placement enables you to co-locate file shares in the identical availability zone as your GPU digital machines, lowering round-trip latency for the data-intensive reads these workloads rely upon. The provisioned v2 billing mannequin enables you to measurement IOPS and throughput independently to maintain GPUs fed. The result’s quicker scaling, greater GPU utilization, decrease value per inference, and a extra responsive expertise in your clients.
Scale cloud-native functions with shared storage
Cloud-native functions typically want shared storage that works instantly with Kubernetes-based deployment fashions. That is particularly necessary for workloads that want ReadWriteMany (RWX) entry, persistent shared state, or storage that evolves with the appliance.
Azure Information integrates with Azure Kubernetes Service via the Azure Information CSI driver, enabling Kubernetes-native workflows corresponding to dynamic provisioning via StorageClasses, expandable persistent volumes, and shared storage throughout a number of pods with ReadWriteMany entry.
The cloud-native story can be about scale and velocity. The brand new file share expertise helps as much as 10,000 file shares per subscription per area, serving to groups help giant multi-tenant or high-density environments. It additionally provisions about 2.5 instances quicker, with even better positive factors in batch deployments, making it simpler to scale shared storage rapidly as utility demand grows. The provisioned v2 mannequin additionally helps groups handle complete value of possession as they scale: you can begin with smaller shares, develop capability incrementally as demand rises, and use percentage-based metrics to observe utilization and right-size shares over time.
Collectively, these capabilities make Azure Information effectively suited to cloud-native situations corresponding to shared content material repositories, configuration and artifact shops, CI/CD workflows, and containerized companies that want persistent file storage throughout replicas or nodes.
Zooniverse, the world’s largest platform for people-powered analysis, places this into observe. After migrating from AWS to a totally managed Azure Kubernetes Service setting, the group runs self-hosted Redis situations backed by Azure Information to supply caching and persevered information storage. The platform helps roughly 100 lively initiatives and 10,000 to fifteen,000 each day customers, with site visitors that may spike to tens of hundreds of API calls per minute. Deployments that after took an hour now run in three to 10 minutes, and the transfer to managed companies on Azure decreased the group’s operational workload by about one full-time engineer.
Modernize enterprise functions with much less disruption
Many enterprise functions constructed on NFS nonetheless rely upon acquainted Linux and POSIX-style behaviors, secure file semantics, and operational continuity. That makes wholesale refactoring costly and dangerous.
Azure Information helps organizations modernize these functions by offering managed NFS file shares for Linux workloads whereas preserving acquainted entry patterns. That makes it simpler to maneuver present Linux line-of-business functions like SAP and different workloads that rely upon POSIX-compliant file shares, case sensitivity, or Unix-style permissions with out forcing groups to refactor them first.
Organizations can use the brand new Azure Storage Mover and Azure Migrate help for NFS to evaluate, plan, and transfer Linux-based file workloads into Azure with much less disruption to present operations. Groups that favor their present tooling also can run third-party migrations from on-premises NAS utilizing companions corresponding to Komprise, giving Linux file estates multiple path into Azure. The result’s a sensible, end-to-end path for bringing present Linux and NFS environments into Azure, the place evaluation, migration, and modernization occur as one movement slightly than a collection of disconnected steps.
As soon as functions are working in Azure, organizations also can strengthen enterprise continuity and safety. Options corresponding to snapshots and tender delete assist enhance restoration choices and shield business-critical information, whereas encryption in transit for NFS helps safe information shifting between purchasers and shares. Mixed with Azure networking choices corresponding to VPN and ExpressRoute, Azure Information offers resilience, and operational basis enterprises anticipate for long-running Linux functions.
The brand new file share expertise additionally simplifies governance, improves value monitoring on the share stage, and permits extra granular community configuration and role-based entry management for various groups and functions.
Medline, an over $25 billion chief in medical-surgical provide manufacturing and distribution, exhibits this modernization path in motion. As a part of its migration from an on-premises SAP ECC on HANA setting to Azure, Medline constructed a cloud-native SAP answer leveraging totally managed Azure Information shares to supply zonally resilient shared storage. The re-architecture improved transaction instances for key SAP workloads by greater than 80 %, reduce response instances for probably the most important transactions by practically 60 %, and accelerated IDoc processing by greater than 50 %.
Construct extra with companions and ISVs
A key cause Azure Information matches accomplice ecosystems is that it speaks customary protocols slightly than a proprietary interface. By exposing industry-standard SMB 3.x and NFS 4.1 endpoints, Azure Information works with present functions, instruments, and frameworks with out code adjustments.
Azure Information NFS additionally expands what companions, ISVs, and platform groups can construct on high of the service. Capabilities such because the NFS Snapshot, Delicate Delete, and REST API for NFS shares assist ecosystem companions combine Azure Information into broader options for BCDR, information motion, administration, and utility enablement.
Within the SaaS and ISV house, Azure Information also can function shared storage behind accomplice functions that want acquainted POSIX-style entry patterns with out forcing clients to handle file infrastructure themselves. These similar protocol and data-protection primitives additionally make Azure Information a goal for database backup workloads—together with SAP and Oracle backups—that companions and ISVs more and more run in opposition to NFS or REST endpoints.
Azure Information additionally matches into adjoining developer and platform situations. For instance, new GitHub-based workflows on Azure Kubernetes Service could use shared storage for runners, caching, artifacts, or job state. These situations present how Azure Information can convey collectively utility platforms, developer tooling, and shared file storage.
Throughout AI, cloud-native, enterprise, and ecosystem situations, Azure Information NFS offers organizations a single managed file platform for contemporary Linux workloads. It helps groups help shared storage wants in Azure with out stitching collectively separate options for every use case. To get began, discover Azure Information documentation. If you want to be taught extra or focus on your state of affairs, contact us at azurefiles@microsoft.com.
