SAN stands for Storage Area Network and like all networks SANs can be configured in a variety of physical and logical layouts referred to as a topology. These networks use a fiber channel fabric which is designed specifically to manage communication between storage devices and servers or workstations.
A typical SAN consists of the following:
- Devices on the edge of the network
- Switches at the core of the network
- Cabling linking the Nodes and Switches together
The topology of a SAN defines how the switches which make up the SAN fabric are interconnected. The goal of a SAN topology is always to optimize functionality, redundancy and scalability. The optimal topology for this is almost always a core-edge topology, however depending on the exact application of the SAN and the costs involved it is not always guaranteed to be the one used.
San Topology Examples:
Star (single switch) - A Star topology consists of a single switch which connects all Nodes in a SAN. This topology is the cheapest and simplest implementation of a SAN, but also has a single point of failure and as a result no redundancy.
Cascade - A Cascade topology is comprised of a series of interconnected switches in a queue. Switches on one end of the queue may not be connected directly together, leading to a large number of switch hops required to transmit information from one edge device to another.
Ring - A Ring topology is made up of multiple switches connected to 2 other switches in the fabric so every device connected to every switch can communicate to eachother. This does mean that devices on different switches may need to travel multiple hops to communicate however. A break in the Ring can still mean devices may communicate, however they may need to travel over more switch hops.
Full Mesh - In a Full Mesh SAN, all switches are connected to all other switches within the fabric. This means that any device on any switch can communicate to any other device with a guarantee it only has to travel over a single switch hop. The major disadvantage with a full mesh is scalability, as you add more switches, more ports on every switch are being taken up to connect them together.
Core-Edge - A Core-Edge topology maintains the guarantee of a single switch hop that a full mesh topology does, without the requirement of interconnecting all switches. A high availability/high performance director is used as the core switch connecting high performance servers and storage devices. Devices with lower performance requirements can be connected through edge switches.
Edge-Core-Edge - An Edge-Core-Edge topology separates server and storage devices onto separate edge environments, leaving the core for switch interconnections exclusively. This enables a SAN to be expanded by storage size or by computing power independently.
Edge-Core-Core-Edge - An Edge-Core-Core-Edge topology is used when connecting two core switches which are a great physical distance apart. The connections between these switches are called a backbone, and there are typically redundant connections put in place between the two core switches.