Expanding Storage? Know Your NAS from Your SAN
SAN (Storage Area Network) solutions have historically targeted large, enterprise-grade consumers. With prices continually dropping, SAN storage is looking more and more attractive to budget-constrained operations. The technology is also changing, and there's some new and exciting terminology to learn. This article's goal is to provide a very light introduction to SAN technologies and bring to light some new options for medium-sized IT shops' storage needs.
Hard drives are cheap, but keeping them in a single file server isn't scalable. It is customary to purchase stand-alone RAID arrays and connect them to a file server via SCSI cards. In this Direct Attached Storage (DAS) configuration, only one server can access the storage. With Network Attached Storage (NAS) and SAN storage, multiple computers can access the storage space, albeit with some restrictions. The end result when using NAS or SAN solutions is increased upgradeability, manageability, and flexibility.
Puzzling Out the Alphabet Soup
Most NAS solutions simply share files using the SMB/CIFS or NFS protocols. NAS devices are basically file servers themselves, so all hosts need to be able to speak the language the NAS device is using. NAS devices are advantageous because multiple servers or hosts can access the same files at the same time, but they certainly aren't the most efficient solution.
SAN devices frequently use fibre channel connections, at 2 Gbps. Storage comes in RAID array boxes with multiple fibre channel connections per controller, normally used to connect them to a SAN switch or directly to file servers. In either configuration, the RAID needs to be configured to assign certain amounts of disk space to specific controllers. With a SAN, you can have 5TB of space, but break it up and assign 1TB to five servers, for example. The actual amount will be less, since you'll first want to use RAID 5 with a hot spare, or something similar.
SAN devices are networked using direct fibre channel connections. These direct connections provide many advantages. First, you aren't contending with other devices on a network, as you would be if using a NAS solution. Second, 2Gbps. And finally, fibre channel Host Bus Adapters (HBAs) are designed with storage specifically in mind. Fiber channel devices normally use SCSI commands to communicate, so they don't have nonessential complexity or overhead.
A newer SAN protocol, iSCSI, is poking its head through the clouds as well. The iSCSI protocol sits on top of TCP and provides the basic SCSI command set. Of course, there's overhead, so demanding applications probably can't use iSCSI right now. Some TCP tweaking makes does it a viable option for some less demanding applications, especially with the advent of 10Gb Ethernet.
Again, SAN switches are quite expensive, and sometimes unnecessary. The fibre channel ports on the storage arrays can connect directly to file servers via a fibre channel HBA. Old SCSI DAS storage was normally just a few hundred gigabytes, and extremely expensive. In many medium-sized outfits 5TB of space would be a significant improvement. SAN storage can now be purchased with serial ATA (SATA) (define) drives that perform better than the SCSI drives of yesteryear. If you're stuck with 500GB SCSI arrays connected to a file server, it might be time to look at SAN storage, without a switch for starters.
SAN switches are just like network switches. They connect to each storage array, and then to each file server as well. The advantage is that you can add new storage and assign it to a specific host port without having to connect new fiber to the file server. Many products even allow the SAN storage arrays to be managed via the same software that's used to manage the switch, making RAID configurations a snap.
Watch for Interop Pitfalls
Be wary about mixing and matching SAN switches and storage arrays though: they rarely interoperate between various vendors. There are standards for SAN device communication, but many companies began offering SAN solutions before these standards were finalized. Currently most vendors' equipment does interoperate, but normally you sacrifice management features if storage arrays are used with unintended SAN switches. SAN switches are unfortunately expensive, and we wanted to mainly talk about the advantages of simple SAN storage here.
SAN storage prices have dropped dramatically with recent hard drive technology advances and increased storage demands. There's still the option of getting fibre channel or SCSI drives, but SATA drives offer quite a benefit as well. SATA is cheap, fast, and reliable. People generally trust SCSI the most and IDE the least, but higher-end SATA drives are trustworthy, and fast. And looking at it a bit more pessimistically, if a drive dies it is really cheap to replace. Assuming you configured the array properly, and only one disk dies at a time, the RAID should be able to rebuild the data from the missing disk.
If your operation desperately needs storage, but is stuck with older SCSI DAS, we'd recommend looking at SAN devices, sans the switch. If you purchase a 5TB array and connect it to multiple file servers, it will help build some confidence in SATA technology, as well as provide a huge increase in space. Later, if space requirements make it necessary, you can always get a SAN switch before purchasing the next SAN array; or not. Switches are great for large environments where you want to allocate disk space to servers without having to run more fiber, but in smaller organizations it might not be worth the expense and added complexity.