With 200TB of needed storage, is a SAN a good idea?

If you can afford it, a centralized storage system is ideal for safely and reliably serving your company's data.

That said, check your math- while it's fine to back servers and computers up to a NAS, it's almost never the best choice. Archives are also not optimal for a NAS. Archives and backups are best put on something else like a tape drive or deduplicated disk pool. Either way, you need something like this to back up the primary data stored on the NAS.


For 200TB, yes you want an array. However you're thinking in the wrong direction if you're trying to square off SAN vs. NAS. They're different tools for different jobs.

NAS is network accessible - typically delivering NFS or CIFS storage, sometimes iSCSI. (And sometimes FCOE).

SAN typically means a fiber channel network providing block storage over SCSI to a host.

Fundamentally, as SAN is a high speed, low latency network dedicated for storage.

In either case though, the array you're talking to is still pretty similar - it's a set of disks configured in a RAID set. (Which configuration you use depends on budget and performance requirement).

So the real question is - how much performance or concurrent access do you need? NFS and CIFS over Ethernet are fine for moderate performance loads (especially over a 10G network) but there's always going to be more protocol overhead than Fiber Channel. However, they're also quite good for sharing storage between multiple hosts/users for concurrent access.

SAN has more cost overhead - a pair (at least) of HBAs per server, and ports on a fabric switch. But it tends to be faster.

Seperate from this is the difference between backup/archive disk profiles and front end storage - front end storage needs good peak performance. Backup doesn't really.


If you have a tight budget, a bit of free time and some geek skills you can also consider using some of the distributed file systems like MooseFS or GlusterFS (more options here: http://en.wikipedia.org/wiki/Comparison_of_distributed_file_systems). I played a bit with MooseFS in my previous work and it served well as a file storage. It won't be as fast as e.g. NetApp but much, much cheaper. And you won't be limited to specific hardware vendor because you can connect anything to DFS to expand the storage (servers, desktops, laptos with any kind of the disks). Of course network bandwidth will be your bottleneck in this case.