By Rick Cook
If a SAN carries most, or all, of a company's storage load, it had better not go down. Redundancy is one of the most straightforward ways of protecting critical parts of any computer system and in relation to SANs, redundancy has two common meanings.
The simplest, cheapest level of redundancy is path redundancy, which provides two separate data paths between storage and servers by including two Host Bus Adapters (HBAs) in each node. The data paths use separate ports in the Fibre Channel fabric switch, which effectively isolates them. The SAN software must support automatic failover from one HBA to the other in the event a data path fails. Redundant data paths are especially suitable for networks where only one or two of the connections need high reliability. It is susceptible to a failure of the fabric switch but otherwise the connection is well protected.
For critical applications you can provide two independent, identical SAN fabrics so that even the failure of a SAN switch will not shut the system down. This solution effectively doubles the cost of the SAN since everything from the HBAs to the fabric switch to the storage arrays must be duplicated. However, it does provide a very high level of reliability.
Of course, high availability also requires protecting and perhaps duplicating other parts of the system, such as the LAN and especially the supply of power.
Dell discusses redundancy in SAN fabrics as well as a number of other important topics in a white paper on designing SAN architectures on its Web site at
While the discussion is built around Dell's SAN offerings, it is sufficiently general to be useful to anyone who is considering implementing a SAN.
About the author: Rick Cook has been writing about mass storage since the days when the term meant an 80K floppy disk. The computers he learned on used ferrite cores and magnetic drums. For the last twenty years he has been a freelance writer specializing in storage and other computer issues.