AZ-305 Practice Question 25: High Performance File Share
Microsoft Azure Solutions Architect Expert (AZ-305) · Test 3 · question 5 of 10
A pharmaceutical company is migrating a genomics pipeline out of its own datacenter. The pipeline's compute nodes run Linux and mount the dataset over NFS, while a Windows reporting farm must read the *same* files over SMB. The design must provide:
- a single 300 TB volume, since the pipeline does not support sharding across shares, - sub-millisecond latency for random reads, which the on-premises filer currently delivers, - volume snapshots with in-place revert, and replication of the volume to a second Azure region, - a fully managed service, not a file server the company operates.
Which storage service should the design use?
Show answer and explanation
Answer:
Azure NetApp Files is the managed NAS service Microsoft positions for demanding, high-performance, low-latency workloads with advanced data management, the ones it describes as otherwise "un-migratable". It is the only option here that meets all four constraints at once: it supports dual-protocol volumes so NFS and SMB clients read the same data, its service levels provide sub-millisecond minimum latency, large volumes scale to 2 PiB so 300 TB is a single volume, and it provides volume snapshots with in-place revert plus cross-region replication.
Option A breaks on two limits. An Azure Files share tops out at 256 TiB, below the 300 TB the pipeline needs in one volume, and Azure Files documents single-millisecond minimum latency rather than sub-millisecond. A share also cannot serve NFS and SMB against the same data, which is why the option has to bolt on an AzCopy job, a second copy that is immediately stale and doubles capacity.
Option B changes the access model. Blob NFS 3.0 gives the Linux nodes a mount, but the reporting farm would have to be rewritten against the REST API instead of an SMB path, and blob storage is object storage: it offers no dual-protocol file semantics and no volume snapshot with in-place revert.
Option C can be made to work technically and fails the last requirement by definition. The company would own the operating system, the NFS and Samba configuration, the disk striping, the patching, and the high availability of a single file server, and it would still have to build its own snapshot and cross-region replication mechanism.