Applies toAzure Blob Storage / Data Lake Storage, Azure Files (Standard, Premium), Azure NetApp Files, managed disks, Azure Elastic SAN
Start from how the data is accessed
| Access | Service |
|---|---|
| Object/REST (apps, data lakes, backups, media) | Blob Storage (with hierarchical namespace for analytics = Data Lake Storage) |
| File share over SMB or NFS, lift-and-shift file servers, shared app config | Azure Files |
| High-performance NFS/SMB with enterprise NAS features (SAP, HPC, Oracle, VDI profiles at scale) | Azure NetApp Files |
| Block storage for one VM (OS, database data files) | Managed disks |
| Block storage pooled for many VMs/clusters over iSCSI, SAN-style consolidation | Elastic SAN |
Comparison
| Blob | Azure Files | Azure NetApp Files | Managed disks | Elastic SAN | |
|---|---|---|---|---|---|
| Protocol | REST, NFS 3.0 (opt.), SFTP (opt.) | SMB, NFS 4.1 (Premium) | NFS 3/4.1, SMB, dual-protocol | Block (attached) | iSCSI |
| Shared by many clients | Yes | Yes | Yes | One VM (shared disks for clusters) | Yes |
| Performance | Massive throughput, object latency | Standard/Premium tiers | Service levels, low latency | Per disk type/size | Pooled, per volume |
| Snapshots / replication | Versioning, object replication, GRS | Share snapshots, GRS (Standard) | Snapshots, cross-region/zone replication | Snapshots, ASR | Snapshots |
| Identity | Entra ID, SAS, keys | AD DS / Entra Kerberos for SMB | AD for SMB, NFS Kerberos | n/a (VM-level) | n/a |
| Cost model | GB by tier + transactions | Provisioned (Premium/v2) or pay-as-you-go | Provisioned capacity pools | Provisioned per disk | Provisioned base + capacity |
Check current feature, protocol and region availability for your choice; these services gain features often.
Common mappings
| Workload | Typical choice |
|---|---|
| Windows file server replacement | Azure Files (+ File Sync for branch caching) |
| FSLogix profiles (AVD) | Azure Files Premium; Azure NetApp Files at large scale |
| SAP HANA shared/data volumes | Azure NetApp Files or certified disks per SAP guidance |
| Data lake for analytics | Data Lake Storage (Blob with hierarchical namespace) |
| Backups, archives, logs | Blob with lifecycle tiers |
| SQL Server on VM | Managed disks (Premium SSD v2 / Premium SSD) |
| Kubernetes persistent volumes | Disks (RWO), Azure Files or NetApp Files (RWX) |
| Many VMs needing pooled block storage | Elastic SAN |
Design questions
- Which protocol does the app actually speak? Rewriting apps from SMB to REST is a project, not a setting.
- How many clients read and write at once?
- What latency and throughput at peak (measure, don't guess)?
- What restore granularity is needed (file, share, volume)?
- How does it replicate for DR, and what RPO does that give?
- How is it reached privately (private endpoints, delegated subnets) and authenticated?
Traps
| Trap | Avoid by |
|---|---|
| SMB port 445 blocked from on-prem/ISP | Plan VPN/ExpressRoute or Azure File Sync |
| Standard file share too slow for an app | Premium Files or NetApp Files; test with real load |
| Using blobs as a file system for legacy apps | Use Files/NFS; BlobFuse has caveats for heavy write patterns |
| NetApp Files capacity pools sized too large | Size by service level and throughput need; resize pools over time |
| Forgetting backup | Azure Backup supports Files, Blob, disks; ANF has its own backup; confirm coverage |
Tip
Run a two-week pilot with real data and real client counts. Storage performance on paper rarely matches small-file metadata-heavy workloads in practice.
Checklist
- Access protocol and sharing model identified
- Performance measured on the current system
- Service chosen and features verified in region
- Identity, private networking and backup designed
- DR replication and RPO confirmed
- Pilot with real workload