Accelerate modern Linux workloads with Azure Files

Linux workloads are progressing rapidly as companies update their on-premises applications, embrace cloud-native setups, and develop AI and data-heavy pipelines in the cloud. To keep pace with this swift transformation, teams require a managed file platform that offers the performance, reliability, security, and adaptability needed for these workloads within Azure.

Azure Files provides an effective solution for this shift, offering fully managed file storage tailored for Linux workloads. It blends familiar file access methods with integrated performance, data protection, security, and seamless integration with Azure services. This means you can easily use shared file storage in the cloud without the hassle of directly managing the underlying infrastructure.

In this article, we’ll look at how Azure Files NFS supports modern Linux workloads across various sectors, including AI, cloud-native applications, enterprise updates, and collaborations with partners or ISVs.

Accelerate AI and Data-Intensive Workloads

AI inferencing involves loading a trained model to deliver predictions. Before any request can be processed, the model’s weights—details of what it learned during training—need to be transferred from storage to the serving instance. This step is time-consuming as model sizes have ballooned to tens or even hundreds of gigabytes. Consequently, how quickly these weights are loaded can significantly influence the responsiveness of an endpoint. It becomes critical especially when a replica starts, and with GPUs being one of the priciest resources, minutes wasted waiting for weights can lead to lost costs.

In the past, teams would either include model weights directly in their container images or have each replica download a copy, leading to oversized images, duplicate data, and slow cold starts just when rapid scaling is essential. With Azure Files, however, the weights are stored once on a file share that all replicas can access and read simultaneously. This approach keeps images smaller, requires managing one version, and allows new replicas to come online in mere seconds for smaller models. Additionally, the Linux NFS nconnect mount option means a client can establish several parallel TCP connections to the same file share, enhancing throughput when scaling up.

Moreover, the new Zonal Placement feature enables you to co-locate file shares with your GPU virtual machines, minimising latency for data-intensive reads necessary for these workloads. With the provisioned v2 billing model, you can independently size IOPS and throughput to ensure GPUs are well-supported, resulting in improved scaling, enhanced GPU usage, reduced inference costs, and a more satisfying user experience.

Expand Cloud-Native Applications Using Shared Storage

Cloud-native applications often require shared storage that works effectively with Kubernetes deployment models. This is especially crucial for workloads needing ReadWriteMany (RWX) access, persistent shared state, or storage that adapts alongside the application.

Azure Files integrates seamlessly with Azure Kubernetes Service via the Azure Files CSI driver. This enables Kubernetes-native processes such as dynamic provisioning through StorageClasses, scalable persistent volumes, and shared storage across multiple pods with RWX access.

Additionally, cloud-native applications thrive on scale and speed. The new file share experience accommodates up to 10,000 file shares per subscription per region, empowering teams to support extensive multi-tenant or high-density setups. It also provisions around 2.5 times faster, particularly benefiting batch deployments that enhance the ease of scaling shared storage in response to application demand. The provisioned v2 model further assists teams in managing costs, allowing initial smaller shares that can be incrementally increased as needed, while employing percentage-based metrics for ongoing resource monitoring and optimal sizing.

These features make Azure Files a perfect choice for cloud-native scenarios like shared content repositories, configuration and artifact stores, CI/CD workflows, and persistent file storage across replicas or nodes in containerised services.

Zooniverse, the largest platform for user-powered research, illustrates this in action. After moving from AWS to a fully managed Azure Kubernetes Service environment, their team can run self-hosted Redis instances backed by Azure Files to allow for caching and data storage. Supporting about 100 active projects and 10,000 to 15,000 daily users, Zooniverse can now handle traffic surges of tens of thousands of API calls per minute. Deployments that previously took an hour are now completed in just three to ten minutes, while transitioning to managed services on Azure has freed up as much workload as one full-time engineer.

Modernise Enterprise Applications with Minimal Disruption

Many enterprise applications that rely on NFS require consistent Linux and POSIX-style behaviour, stable file semantics, and operational continuity. This can make extensive refactoring both costly and risky.

Azure Files assists organizations in modernising these applications by providing managed NFS file shares tailored for Linux workloads while maintaining familiar access patterns. This allows seamless migration of existing Linux business applications like SAP, or others that rely on POSIX-compliant file shares, case sensitivity, and Unix-style permissions without necessitating extensive refactoring first.

Leverage the new Azure Storage Mover and Azure Migrate support for NFS to evaluate, plan, and transition Linux-based file workloads into Azure with minimal interruption to existing operations. For teams who prefer their current tools, third-party migrations from on-premises NAS can be run through partners like Komprise, providing more than one route for migrating Linux file environments into Azure. This offers a straightforward, cohesive approach to bringing existing Linux and NFS setups onto Azure, ensuring assessment, migration, and modernisation occur smoothly and not as disconnected steps.

After your applications are operational on Azure, it’s also possible to enhance business continuity and security. Features such as snapshots and soft delete improve recovery options and secure crucial data. Furthermore, encryption in transit for NFS ensures that data remains secure during transfers between clients and shares. When combined with Azure’s networking options like VPN and ExpressRoute, Azure Files presents the resilient operational foundation that businesses expect for long-standing Linux applications.

The new file share experience also streamlines governance, enhances cost tracking at the share level, and allows for more detailed network configurations and role-based access control tailored for different teams and applications.

Medline, a medical-surgical supply leader with over $25 billion in revenue, exemplifies this modernization journey. As part of transitioning from an on-premises SAP ECC on HANA environment to Azure, Medline developed a cloud-native SAP solution employing fully managed Azure Files shares for zonally resilient shared storage. This re-architecture led to an impressive 80% improvement in transaction times for vital SAP workloads, nearly 60% faster response times for critical transactions, and over 50% acceleration in IDoc processing.

Collaborate More Effectively with Partners and ISVs

One of the main advantages of Azure Files is its compatibility with partner ecosystems, largely due to its use of standard protocols instead of proprietary interfaces. By providing industry-standard SMB 3.x and NFS 4.1 endpoints, Azure Files supports existing applications, tools, and frameworks without requiring any code alterations.

Azure Files NFS also broadens what partners, ISVs, and platform teams can create using the service. Features like NFS Snapshot, Soft Delete, and the REST API for NFS shares enable ecosystem partners to integrate Azure Files into wider solutions for business continuity, data movement, management, and application enhancement.

In the SaaS and ISV sector, Azure Files can be utilized as shared storage for partner applications that require familiar POSIX-style access without imposing file infrastructure management on customers. These same data protection features also make Azure Files an ideal choice for database backup operations—including those for SAP or Oracle—run against NFS or REST endpoints.

Azure Files fits seamlessly into various developer and platform scenarios. For instance, new GitHub-based workflows on Azure Kubernetes Service may require shared storage for runners, caching, artifacts, or job states. These instances highlight how Azure Files can integrate application platforms, developer tools, and shared file storage.

For a range of sectors like AI, cloud-native applications, enterprises, and collaborations, Azure Files NFS provides organisations with a single managed file platform tailored for modern Linux workloads. It helps teams meet their shared storage needs in Azure without having to piece together separate solutions for each requirement. To get started, check out the Azure Files documentation. If you’re keen to learn more or discuss your specific needs, feel free to reach out at [email protected].