Understanding the Three Core Storage Architectures: DAS, NAS, and SAN

A colleague in operations complained about choosing the wrong storage architecture, which prompted the author to explain why DAS, NAS and SAN are unavoidable choices for any company dealing with data storage.

DAS: Direct‑Attached Storage

Characteristics : Storage devices connect to a single server via SCSI, SAS or similar interfaces, forming a private link without any network equipment. From the OS perspective the devices appear as local disks.

Advantages :

Very low latency because data bypasses network stacks; SAS 12 Gb/s channel is dedicated to the server.

Predictable performance without network‑induced jitter.

Relatively low cost – no extra networking hardware or management software.

Low technical barrier – ordinary system administrators can deploy and maintain it.

Limitations :

Almost no scalability – the number of disks is limited by the server’s backplane.

Data‑island problem – storage attached to one server cannot be shared with others.

Low resource utilization across multiple workloads.

Availability risk – if the server fails, its storage becomes inaccessible.

Long maintenance windows – any expansion requires downtime.

Typical Scenarios :

Small‑business file services (dozens of users).

Development and testing environments where quick local disks are needed.

Single‑node high‑performance computing that cannot tolerate network latency.

Edge‑computing nodes with unreliable network connectivity.

Video surveillance NVRs that require stable bandwidth for continuous recording.

NAS: Network‑Attached Storage

NAS is a dedicated file‑storage appliance that provides file‑level access over standard Ethernet using protocols such as NFS, SMB/CIFS and AFP. The device runs a lightweight OS and manages the underlying file system.

Characteristics : Uses IP networking as the transport layer; clients mount shared folders and operate on files rather than raw blocks.

Advantages :

Strong data‑sharing capability – hundreds of devices can access the same storage simultaneously.

Highly flexible deployment – any location with network connectivity works.

Scalable – capacity can be expanded by adding disk enclosures; performance can be upgraded by swapping to faster models.

Centralized management – quotas, snapshots and backup policies are configured from a single interface.

Cost‑friendly – entry‑level units cost a few ten‑thousands of yuan, high‑end models only a few hundred thousand, far cheaper than SAN.

Limitations :

Performance limited by network quality; 1 GbE rarely exceeds 120 MB/s.

Protocol overhead – each file operation passes through the network and file‑protocol stacks.

Small‑file I/O suffers significant IOPS degradation.

Strong network dependency – switch failures or broadcast storms can interrupt storage access.

Security concerns – data travels over the network and requires VLAN isolation and encryption.

Typical Scenarios :

Enterprise file sharing for hundreds of users.

Media production where multiple editors work on 4K assets; 10 GbE NAS provides sufficient bandwidth.

Virtual machine image storage for VMware or KVM clusters.

Centralized video‑surveillance storage.

Medical imaging (PACS) where large DICOM files need scalable storage.

Code repositories (GitLab, SVN) hosted on NAS for multi‑node access and backup.

SAN: Storage Area Network

SAN connects storage devices to many servers via a dedicated network, offering block‑level services. Servers use HBA cards to attach to the SAN; protocols include Fibre Channel (FC), iSCSI and NVMe‑over‑Fabrics.

Characteristics :

Independent storage network isolated from the business LAN.

Provides block‑level storage; servers create file systems on LUNs.

Multi‑path access for redundancy and bandwidth aggregation.

Storage pooling – multiple arrays form a unified pool that can be allocated on demand.

Advantages :

High performance – FC SAN starts at 16 Gb/s with microsecond‑level latency.

Strong scalability – a SAN domain can serve hundreds of servers and arrays.

High resource utilization – storage is partitioned and allocated as needed.

Rich enterprise features – snapshots, cloning, deduplication, compression, automated tiering.

Robust reliability – dual‑controller architecture, multi‑path redundancy and online firmware upgrades ensure continuous operation.

Limitations :

High cost – FC switches and HBAs can cost more than an entire NAS solution.

Complexity – zoning, LUN mapping and multi‑path software require specialized knowledge.

Vendor‑specific maintenance – troubleshooting fibre‑channel issues often needs OEM support.

Long deployment cycles – planning, implementation and tuning may take weeks.

Stringent infrastructure requirements – fibre cables must avoid tight bends and data‑center environments must meet strict standards.

Typical Scenarios :

Core database storage for Oracle RAC or SQL Server clusters.

Large‑scale server virtualization (vSphere, Hyper‑V) where many VMs share high‑performance disks.

High‑performance computing clusters with massive small‑file I/O.

Financial trading systems where microsecond latency is mandatory.

Hospital PACS core storage for CT/MRI imaging.

Enterprise ERP systems requiring 99.9999 % availability.

The author also mentions FineDataLink , a data‑integration platform that can seamlessly connect to various SAN systems (EMC, NetApp, Huawei, etc.), provides visual monitoring of throughput, latency and error rates, and includes data‑quality verification to ensure source‑target consistency after migration.

Conclusion

Choosing a storage architecture is essentially a trade‑off among performance, cost and complexity. DAS fits small, low‑cost scenarios; NAS excels at file sharing and flexibility; SAN targets high‑performance, enterprise‑critical workloads. There is no universally best solution—only the most appropriate one for the given business requirements.

Understanding these three architectures helps avoid costly mis‑steps and reduces overtime during storage planning.