What Are Storage Area Networks?

Storage area networks (SANs) are dedicated high-speed networks that connect servers to block-level storage devices, making shared storage resources appear to each server as locally attached disk. Unlike file-based network-attached storage (NAS), which operates at the file system layer over standard Ethernet, a SAN operates at the block level, allowing multiple servers to directly read and write raw storage blocks on centralized disk arrays. This architecture emerged in the late 1990s alongside the growth of enterprise databases and transaction processing systems, where low latency, high throughput, and guaranteed block delivery became requirements that general-purpose local area networks could not consistently meet.

SAN Architecture and Protocols

A storage area network consists of three main components: host bus adapters (HBAs) installed in servers, a dedicated switching fabric, and storage arrays. The fabric routes block-level I/O traffic between initiators (server HBAs) and targets (storage ports) using one of several transport protocols. Fibre Channel (FC), standardized under the T11 technical committee and widely deployed in enterprise data centers, provides point-to-point, switched, and loop topologies with throughput rates of 8, 16, 32, or 128 Gbps and lossless, in-order delivery, which makes it well-suited for mission-critical databases and financial transaction systems. Internet SCSI (iSCSI), defined in IETF RFC 3720, encapsulates SCSI commands within TCP/IP packets and transmits them over standard Ethernet infrastructure; because it reuses existing Ethernet switches and standard network interface cards, iSCSI substantially reduces deployment cost and complexity compared to Fibre Channel. Fibre Channel over Ethernet (FCoE) is a third option that encapsulates FC frames over lossless Ethernet, consolidating storage and data traffic onto a single physical network while preserving the FC management model.

Buffer Storage in SANs

Buffer storage within a SAN serves two distinct purposes. At the device level, storage arrays contain large DRAM and solid-state caches that absorb bursty write traffic, aggregate sequential reads, and mask the latency difference between spinning-disk media and host expectations. This cache is the primary mechanism by which disk-based arrays sustain the consistent sub-millisecond response times that database workloads require. At the fabric level, Fibre Channel switches implement credit-based flow control, where each port maintains a buffer credit count that limits how many frames can be in flight simultaneously, preventing buffer overflow and packet loss. The interplay between device-level buffering and fabric-level flow control is central to SAN performance tuning, as surveyed in research on storage class memory published on arXiv, which documents how emerging non-volatile memory technologies are reshaping the assumptions that traditional SAN buffer design relies upon.

SAN Management and Integration

SANs are managed through fabric management software that handles zoning (controlling which initiators can communicate with which targets), LUN masking (controlling which storage volumes each server can see), and path management (balancing traffic across multiple physical connections for redundancy and throughput). SANs differ from local area networks not just in protocol but in operational philosophy: a LAN is generally configured to be open and discoverable, while a SAN is configured to be isolated and access-controlled by default. As documented in NIST SP 800-209, Security Guidelines for Storage Infrastructure, storage networks require specific security controls distinct from general network security because a misconfigured SAN fabric can expose an entire organization's data to any attached host.

Applications

Storage area networks have applications in a range of enterprise and institutional settings, including:

• Relational database hosting: providing consistent low-latency block I/O for OLTP systems
• Server virtualization: shared storage enabling live migration of virtual machines between hosts
• Disaster recovery: synchronous and asynchronous replication between geographically separated SANs
• High-performance computing: parallel access to shared datasets by many compute nodes simultaneously
• Backup and archiving: tiered storage pools that migrate cold data from fast disk to tape