openskills.info
Course catalog
Course Preview

Storage Area Networks

A storage area network carries block-storage traffic between host initiators and storage targets. The fabric, endpoint identities, zoning, logical-unit masking, and multipath software determine which hosts see each remote block device and how paths survive failures.

itStorage, backup, and data protection

Don't Panic — Storage Area Networks

A storage area network (SAN) is a dedicated network whose only job is to carry disk traffic between servers and storage systems. The server sees something that looks like a local hard drive. The actual blocks live on a storage array somewhere across the room, reached through a switch.

Before SANs, storage was mostly bolted inside each server. If one machine filled up while its neighbour had space to spare, that was simply too bad. A SAN pools the capacity and hands slices of it out over the network, so one array can serve many hosts and you can grow or reassign storage without opening a case.

Three ideas carry the whole topic. The first is the SCSI model of roles: an initiator, which is the server, sends storage commands; a target, which is the array, receives them; and a logical unit, addressed by a number called a LUN, is the block device the target hands back.

The second is that names are deliberately permanent. Each port carries a fixed identifier — a worldwide name on Fibre Channel, an iSCSI qualified name on iSCSI — because the physical adapter and its network address can change, and the rules that grant access must not move when they do.

The third is that access control has two separate layers that both have to agree. Zoning lives in the fabric and decides which ports are allowed to talk to each other at all. LUN masking lives on the array and decides which initiators it will actually present a given logical unit to. You configure both, every time.

The surprise for most people is that presenting one logical unit to two servers does not give you shared storage. The transport moves blocks; it never coordinates who writes them. Put an ordinary file system on that device, mount it on two hosts, and they will quietly corrupt each other. Genuinely shared access needs a cluster-aware file system or an application that arranges the coordination itself.

The other trap is multipathing. You run two cables for redundancy and software merges them into one device. But two cables through a single switch, or into a single array controller, still fail together. Independent paths mean separate hardware the whole way from host to array.

For the rest of the course: the Intro gives you the full component map and the data path; the Slides lay the transports and access layers out side by side; the Cheatsheet is what you want open while tracing a request or choosing a control; and the Reference tab points at the standards and vendor documentation for exact procedures. The habit that pays off everywhere is following one request through every layer — host, fabric, target, array — and asking which layer owns what.

Where this skill leads

Relevant careers

See how this topic contributes to broader role-level skill maps.

Sources