Azure Native PowerScale: Dell Managed File Service in Azure Marketplace Preview

Dell and Microsoft have moved PowerScale’s OneFS into the Azure control plane with a Dell‑managed, Azure‑native offering that has entered public preview — a change that converts PowerScale from a self‑deployed cluster inside a customer’s subscription into a transactable, Dell‑operated file service visible and consumable through the Azure Portal and Marketplace.

Background: why this matters now​

The cloud storage landscape is bifurcating: organizations either re‑platform to native cloud storage models or bring enterprise storage features into cloud as managed services. Dell’s public preview of an Azure‑native PowerScale (marketed under the APEX File Storage / APEX File Storage for Microsoft Azure umbrella) addresses the latter path by delivering OneFS capabilities — global namespace, multi‑protocol access (NFS, SMB, S3, HDFS), snapshots, deduplication, SmartLock immutability, and CloudPools tiering — as a managed service integrated into Azure tooling and procurement channels. This is significant because many enterprises running large unstructured datasets (AI/ML training sets, genomics, EDA, HPC, media processing) have relied on PowerScale’s scale‑out OneFS on‑premises. Offering the same feature set as an Azure Marketplace, transactable service reduces operational overhead and can accelerate adoption of Azure AI and GPU compute without wholesale data re‑architecture. Multiple Dell communications and independent reports confirm the public preview timing and the strategic tie‑ins to Azure AI tooling.

---

Overview of the new Azure‑native PowerScale offering​

Two deployment models — choice by design​

Dell now surfaces two clear deployment models for PowerScale in Azure:

• Customer‑managed PowerScale for Azure — the established model where customers deploy OneFS on Azure VMs, manage the VM fleet, networking, OneFS lifecycle, and monitoring themselves. This model preserves maximum control and customization.
• Dell‑managed (Azure‑native) PowerScale / APEX File Storage for Microsoft Azure — Dell provisions and operates the underlying infrastructure and OneFS software, while customers consume a filesystem resource directly from the Azure Portal. Procurement, billing, and role‑based access integrate with Azure Marketplace and customers’ existing Azure agreements.

This choice matters: the managed path trades deep infrastructure control for simplified operations, while the customer‑managed path keeps full custodian control over configuration, network topology, and lifecycle.

Native Azure integration and operational model​

Implementing PowerScale as an Azure Native ISV integration means concrete changes for administrators and procurement teams:

• Visibility and lifecycle operations surface in the Azure Portal and can be automated with Azure CLI and PowerShell.
• Billing flows through Azure subscriptions and Marketplace offers, enabling use of enterprise discounts or private Marketplace offers when supported.
• Security and governance integrate with Azure role‑based access control (RBAC), Azure Monitor metrics, and Azure billing pipelines.

Behind the scenes, Dell remains the primary platform operator in the managed model; support and escalation paths consolidate to Dell for platform concerns while customers retain responsibility for application integration and access controls.

---

Technical capabilities and verified limits​

OneFS feature parity in the cloud​

The managed Azure variant preserves the core OneFS data services that enterprises depend on:

• Single global namespace and scale‑out filesystem architecture
• Multi‑protocol access (NFS v3/v4, SMB3, S3, HDFS)
• Data services: snapshots, inline compression, SmartDedupe, SmartQuotas, SmartConnect client load balancing, SmartLock immutability, and SyncIQ replication for hybrid disaster recovery
• CloudPools policy‑driven tiering to push cold data to Azure Blob while leaving metadata and smartlinks on the filesystem for seamless hybrid workflows.

These capabilities are material for customers seeking to keep the same application semantics and data services in cloud as on‑premises.

Capacity and scale: what’s verified and what isn’t​

Dell’s official technical documentation for PowerScale on Azure states guidance for the customer‑managed Azure edition of up to 18 nodes and 5.6 PiB of usable cluster capacity per cluster. That guidance appears in Dell product and admin pages and is consistent with Dell’s solution brief for cloud deployments. At the same time, several media reports and secondary coverage have cited a larger 8.4‑petabyte single‑namespace figure for what the managed, Azure‑native edition might support. That specific 8.4‑PB number could not be confirmed in Dell’s public product pages or Microsoft marketplace materials at the time of reporting and should be treated as unverified until Dell or Microsoft publish explicit capacity limits for the Dell‑managed edition. Enterprises must request written capacity guarantees for any production contract.

Instance types, throughput and VM support​

Dell’s OneFS documentation and APEX File Storage admin guides list supported Azure VM families and recommended instance sizes for APEX File Storage deployments (Dds_v5, Eds_v5 families and similar configurations). The admin guide provides specific instance‑level throughput and vCPU/memory combinations and describes how cluster performance scales with node count. Those configuration choices feed directly into planning throughput, IOPS, and capacity.

---

Strengths: what the public preview brings to enterprise IT​

• Operational simplicity and faster time‑to‑value. Shifting infrastructure lifecycle management to Dell reduces internal lift for provisioning, patching, and platform monitoring when compared with a self‑managed Azure VM deployment. The Azure Portal integration can accelerate pilots and procurement cycles.
• Preservation of enterprise data services. OneFS’s advanced features (dedupe, snapshots, SmartLock, CloudPools tiering) remain available in cloud editions, enabling parity with on‑prem production workflows and regulatory controls.
• Azure purchasing and billing alignment. Delivering PowerScale as an Azure Marketplace transactable service means customers can use existing Azure committed spend, enterprise agreements, and Marketplace private offers to provision and manage billing centrally.
• Hybrid continuity for large file estates. SyncIQ and CloudPools enable phased migration, disaster recovery, and cloud bursting, supporting hybrid strategies where data remains partly on‑prem and partly in Azure. This is crucial for workloads that cannot immediately re‑platform to native object semantics.
• Targeting AI/ML and throughput‑heavy workloads. The architecture and instance selections are aimed at high concurrency, large file throughput, and low‑latency access for training pipelines. When paired with Azure GPU compute nodes, PowerScale can act as a high‑throughput data plane for model training.

---

Risks, caveats, and what enterprises must validate​

The managed, Azure‑native model offers convenience, but it raises several important considerations that must be validated during preview tests and procurement negotiations.

1. Capacity guarantees and variability in published figures​

Published capacity numbers for customer‑managed and Dell‑managed editions vary across sources; do not rely on media summaries for hard limits. Enterprises must obtain contractual capacity figures, node counts, and performance SLAs directly from Dell or via the Azure Marketplace offer documentation. The 8.4‑PB single‑namespace claim cited in some coverage remains unverified and should not be treated as contractual.

2. Vendor lock‑in and data gravity​

Moving petabytes of unstructured data into a Dell‑managed tenant or a tightly coupled managed appliance increases data gravity and can complicate future migrations or multi‑cloud strategies. Evaluate exit procedures, supported data export tools, and the cost and performance profile for bulk egress.

3. Cost modeling — storage, network, and tiering​

Cloud economics for file services can be complex: snapshot retention, replication, cross‑region replication, and egress costs can materially change a TCO model. For AI workloads that frequently access data from compute in different zones or regions, run detailed cost modeling that includes ingress/egress, tiered storage movements to Azure Blob, and snapshot retention policies.

4. Latency and performance predictability​

Public cloud network paths, VM host variability, and noisy‑neighbor effects can affect throughput and latency. For HPC and tightly synchronous training workloads, validate end‑to‑end performance under realistic, high‑concurrency loads rather than relying on microbenchmarks.

5. Security, compliance, and control surface​

A managed service introduces a third‑party control surface that must be reconciled with internal compliance regimes. Important questions include:

• Where does telemetry and backup telemetry live (customer tenancy vs Dell tenancy)?
• What are data residency guarantees and regional placement controls during preview and GA?
• How are encryption keys managed — customer‑managed keys (CMK) or Dell keys? What audit trails and access logs are available?
• What are RTO/RPO commitments and operational SLAs for the managed platform?

These must be answered and documented in service agreements before production adoption.

6. Network topology and delegated subnets​

Azure Native integrations typically require delegated subnets and resource provider access. That means network design (NSGs, peering, routing, ExpressRoute/Private Link) needs careful planning to match security posture and performance requirements. Confirm supported network designs in the preview documentation.

---

Practical checklist for evaluating the public preview​

• Confirm the exact edition in scope (customer‑managed vs Dell‑managed) and map operational responsibilities for patching, monitoring, and upgrades.
• Request and document published capacity numbers, node counts, and throughput limits for the managed edition; obtain written guarantees where capacity is a gating factor.
• Validate region availability for the preview and planned GA; confirm data residency and telemetry residency rules.
• Run targeted proof‑of‑concept (PoC) tests for representative workloads (AI training, HPC file I/O, analytics), measuring throughput, concurrency, and end‑to‑end latency.
• Build a complete cost model that includes storage, snapshot frequency, replication, network egress, and CloudPools archive costs to Azure Blob.
• Confirm backup, DR, and replication behavior with SyncIQ and CloudPools and how these integrate with existing on‑prem backup tooling.
• Validate security, auditability, and key management details, and confirm SLAs and escalation paths with Dell.

---

Market and strategic implications​

Delivering PowerScale as an Azure Marketplace‑transactable managed service continues an industry trend: hyperscalers and enterprise vendors are making feature‑rich, enterprise storage services available natively in cloud portals. For Azure‑centric enterprises, this reduces friction for moving large datasets into cloud compute and AI services. For storage vendors, it offers a new go‑to‑market route that can leverage customers’ existing Azure procurement commitments. Dell’s messaging and press communications position the offering to work tightly with Azure AI tooling and GPU compute, reflecting a strategic push to make PowerScale a first‑class data plane for cloud AI workloads. At the procurement level, marketplace transactions change partner/reseller dynamics and internal purchasing workflows; organizations must decide whether to treat the managed PowerScale resource as a line‑item in their Azure subscription or to maintain traditional vendor contracts outside of Azure billing. Both approaches have trade‑offs in discounts, billing transparency, and contract terms.

---

Where the preview fits into long‑term adoption plans​

This Azure‑native path lowers the barrier to entry for organizations that want enterprise file services without a long on‑prem to cloud migration project. For greenfield or cloud‑first analytics and AI initiatives, the managed PowerScale preview can be a low‑friction on‑ramp. For regulated industries or companies with strict network controls, the customer‑managed edition will remain essential until managed services demonstrate necessary controls, region flexibility, and contractual guarantees.
Given the mix of convenience and potential lock‑in, the prudent enterprise approach is a staged migration:

• Use the preview to validate integration with Azure AI services, test peak throughput, and confirm how CloudPools tiering affects cost and access patterns.
• Start with non‑mission‑critical datasets or hybrid DR contexts to evaluate operational workflows and support responsiveness.
• Only expand to production datasets after contractual capacity and SLA terms are satisfactorily negotiated and after PoC performance baselines meet business needs.

---

Conclusion — a practical assessment​

The public preview of Dell’s Azure‑native PowerScale marks an important evolution in enterprise file storage for Azure: it brings OneFS’s enterprise capabilities into the Azure Portal as a Dell‑operated, transactable resource that reduces operational overhead and aligns with Azure purchasing flows. Dell’s official materials and multiple independent outlets confirm the offering’s public preview timing and the strategy to pair PowerScale with Azure AI compute. This is a welcome option for organizations that need scale‑out file semantics and enterprise data services without redesigning applications for object storage. The offering’s strengths — familiar OneFS features, Azure Portal and Marketplace integration, and hybrid continuity via CloudPools and SyncIQ — are meaningful for AI, HPC, and large data analytics workloads.
At the same time, the managed model introduces important procurement, compliance, and exit‑strategy questions. Capacity figures and limits differ across sources; the 5.6 PiB guidance for the customer‑managed Azure edition is documented in Dell product pages, while larger single‑namespace numbers reported elsewhere remain unverified and should not be used for planning without written confirmation from Dell. Enterprises must run thorough PoCs, obtain contractual capacity/SLAs, and model costs for storage, snapshots, and network egress before moving production workloads. The preview is the right venue to test operational assumptions: confirm region support, measure real‑world performance with target workloads, validate security and key management, and walk contractual exit scenarios with Dell and Microsoft. For teams that succeed in these pilots, Azure‑native PowerScale has the potential to make enterprise‑grade file services a first‑class citizen inside Azure’s AI and compute ecosystem.

---

Source: Neowin Public Preview of Azure Native Dell PowerScale now available