A storage area network (SAN) is a dedicated, high-speed network that connects servers to shared storage devices and presents that storage to each server as if it were a local disk. Because it works at the block level, the lowest level at which operating systems read and write data, servers can format SAN volumes with their own file systems and run databases, virtual machines and other demanding workloads on them. The term covers both the network (switches, adapters and cabling) and, in everyday use, the storage arrays attached to it.
At a glance
- A SAN gives servers block-level access to shared storage, which they see as local disks.
- It traditionally uses a dedicated network, either Fibre Channel or Ethernet with protocols such as iSCSI or NVMe over TCP.
- It is a long-standing foundation for databases, virtualization clusters and other workloads that need fast, shared, resilient storage.
- Storage arrays on a SAN commonly provide snapshots, replication, thin provisioning and redundancy.
- Alternatives include hyper-converged infrastructure, network-attached storage for file sharing, and cloud block storage.
What problem it solves
Storage installed inside individual servers is hard to share, hard to protect and wasteful: one server runs out of space while another has plenty spare, and if a server fails, its data is stuck inside it. Virtualization makes this worse, because moving or restarting virtual machines between hosts is simplest when the hosts share the same storage.
A SAN pools storage in central arrays and makes it available to many servers over a dedicated network. Capacity can be allocated where it is needed, data is protected by the array’s redundancy, and a hypervisor cluster can share volumes so virtual machines can move or restart on another host. Keeping storage traffic on its own network also avoids competition with user traffic.
How it works
Storage arrays. Arrays hold many disks or flash drives, combine them for redundancy and performance, and carve them into volumes (often called LUNs) assigned to specific servers. Controllers in the array handle caching, snapshots, replication and failover between paths.
Network. Servers connect through host adapters to SAN switches, which connect to the arrays. Production SANs typically use two separate fabrics with every server and array connected to both, so that a single switch or cable failure should not cut access. Access is restricted with zoning and masking so each server sees only its own volumes.
Protocols. Fibre Channel is a dedicated storage networking technology. iSCSI carries block storage commands over standard Ethernet and IP networks. NVMe over Fabrics, including NVMe over TCP and over Fibre Channel, is a newer family of protocols designed for flash storage. Choice depends on performance needs, existing skills and budget.
Data services. Arrays often provide snapshots, replication to a second site for disaster recovery, deduplication, compression and encryption, which can affect licensing and pricing.
Our file and object storage solution page covers how block, file and object storage fit together in a storage plan.
When it matters for buyers
- When an array reaches end of support. Renewal quotes for older arrays can be high; compare a new array, hyper-converged infrastructure, hosted storage and storage as a service (STaaS).
- When virtualized or database workloads slow down. Storage latency is a common bottleneck; flash arrays and newer protocols may help.
- When planning disaster recovery. Array-based replication to a second site is a common option, with its own licensing and network requirements.
- When changing virtualization platforms. Confirm the new hypervisor supports your SAN, its protocols and its data services.
- When designing for high availability (HA). Dual fabrics, redundant controllers and multipathing are what keep storage available through component failures.
Questions to ask vendors
- What usable capacity and performance will we get after redundancy, and what assumptions about data reduction are included?
- Which protocols do you support (Fibre Channel, iSCSI, NVMe over Fabrics), and what network equipment is required?
- What data services are included, and which are licensed separately?
- How are controller and component failures handled, and what is the hardware replacement commitment?
- How do support and maintenance costs change after the initial term?
- What are the options for expanding capacity and for migrating data when we eventually replace the array?
- Is the array supported by our hypervisor and backup software versions?
How it differs from network-attached storage (NAS)
A SAN and network-attached storage (NAS) both put storage on a network, but they share it at different levels. A SAN provides raw block storage that a server formats and controls as its own disk, usually over a dedicated storage network. A NAS device runs its own file system and shares files and folders to many users and servers over the ordinary network, using file-sharing protocols such as SMB or NFS. SANs suit databases and virtual machine storage; NAS suits shared documents and home folders. The ways data is organized above the block level, as files or objects, are covered under file systems and object storage.