Qumulo has announced three updates for customers running file and data workloads on Microsoft Azure: an archive tier for Azure Native Qumulo, a Blob Accelerator for accessing existing Azure Blob data, and support for Microsoft’s Cobalt-based instances in Cloud Native Qumulo. The announcements were made at Microsoft Ignite.
The changes address different parts of cloud data management. Azure Native Qumulo (ANQ) gains a way to retain file-system access to data placed in Azure’s offline archive tier. The Blob Accelerator adds a file-access and caching layer for data that remains in a customer’s Blob container. Cloud Native Qumulo (CNQ), meanwhile, can now run on Azure’s ARM-based Cobalt processors.
ANQ Frozen keeps archived files discoverable
ANQ Frozen extends Qumulo’s managed, scale-out file system service to Azure Blob Storage’s Archive tier. Archive is designed for rarely accessed data and offers the lowest storage costs among Azure Blob’s five tiers: hot, cool, cold, archive and smart. It is an offline tier, however, and files must be rehydrated before they can be read or changed. Retrieval can take hours and carries additional costs.
With Frozen, files move from ANQ to Archive as whole files through manual actions or API commands. The files retain their metadata, paths and identity and access attributes. Qumulo says the ANQ namespace remains available for browsing over SMB and NFS, while Qumulo NeuralSearch can search file-system metadata and application information extracted from file headers. That lets users locate likely candidates before requesting retrieval, rather than restoring an entire archive to find a particular item.
Administrators can approve retrieval requests. Files return to a hot tier within Azure’s retrieval window, which Qumulo says is typically two to 12 hours. Existing ANQ Hot and Cold customers can add Frozen while keeping their current paths, permissions and management approach; organizations can also deploy a dedicated ANQ Frozen system. Each system stores data in a dedicated Azure storage account as native objects, which Qumulo says remain accessible through standard Azure tools without Qumulo software in the access path.
Frozen follows Qumulo’s 2024 introduction of ANQ Cold, which uses Azure Blob’s online Cold tier. Frozen reaches the deeper Archive tier, with its longer retrieval process. The new feature is in controlled preview on Azure, with general availability planned for May 2027. Customers can request preview access through their Qumulo account team.
Blob Accelerator works against existing containers
Qumulo Blob Accelerator is intended to connect file-based applications and AI pipelines to data already held in Azure Blob Storage without migrating or copying it. The customer’s Blob container remains the authoritative copy. Applications can access its data through NFS or SMB, while existing applications using Blob APIs can continue to use the container.
Qumulo says the accelerator reduces object-access latency from a stated range of 10 to 150 milliseconds to between one and five milliseconds. Its NeuralCache serves more than 90 percent of repeat reads from NVMe and DRAM, with cached data returned in under a millisecond. Qumulo also says the accelerator eliminates transaction costs after the first read of a given object, addressing API charges associated with repeated reads. These performance and cost figures are company claims.
The same Blob container can be accessed from Qumulo systems in other Azure regions, other clouds, on-premises data centers or at the edge, according to the company. Qumulo describes the access as strictly consistent and says it does not require additional copies. Existing ANQ customers can add the accelerator to a cluster and connect Azure Storage Accounts; new customers can deploy it on their own Azure instances. Qumulo identifies AI training and inference, analytics, media delivery, genomics, and seismic and subsurface imaging as potential workloads.
Qumulo is also a launch partner for Microsoft Azure Blob Access Points. The feature enables Azure services including Microsoft Fabric, Azure AI Foundry and Azure OpenAI to access unstructured file data on Qumulo through Blob Storage APIs. Qumulo says this can be done without copying or migrating the data or building custom integrations.
CNQ adds Azure Cobalt instances
Cloud Native Qumulo now runs on Microsoft Azure Cobalt, an ARM-based processor platform. The company says every CNQ tier, along with Qumulo AI Accelerators and the Blob Accelerator, runs on Cobalt-based instances. Qumulo reports that these instances cost 9 to 48 percent less than comparable x86 instances and are easier to obtain. It says the software, features and single namespace are unchanged, so customers do not need to refactor applications.
ANQ and CNQ differ in how they are operated: ANQ is a fully managed service co-developed with Microsoft, with Qumulo managing the infrastructure. CNQ is deployed in a customer’s Azure subscription, where the customer manages resources such as virtual machines, networking and scaling. Cobalt support applies to the latter service.