Within Reach of Most IT Budgets, Storage Area Networks Pay Off
Products mentioned in this article
Storage area network (SAN) technology has been around since the late 1990s, but historically it was priced far out of the range of small IT departments. Now this option for making more-efficient use of data storage has become affordable for all but the smallest of IT shops.
For example, Internet Small Computer System Interface (iSCSI) SAN arrays such as the Drobo Elite cost $3500, and the Dell PowerVault MD3000i starts at just over $5000.
SAN devices are disk arrays located on a network storage device shared with multiple servers--up to 16 servers, for these models. To each server operating system, however, the storage that a SAN device provides appears to be dedicated, not shared.
It used to be that when you bought a server, you guessed how much storage you would need, and then bought disks based on that often-underestimated number. When your needs outgrew your disk capacity, you had to back up the server, replace the disks, and restore the data.
With a storage area network, however, you buy only as much storage as you initially need, and then expand as you go. This save your company money in hardware costs and energy consumption, and it reduces IT personnel workload and system downtime.
For a small business, high-throughput, low-latency fiber-channel-based SANs are still too pricey. But less-expensive iSCSI SANs, which use standard ethernet cables, network cards, and switches, have no trouble keeping up with the disks that small and medium-size companies are likely to use.
iSCSI, a protocol for connecting storage devices using the IP protocol and SCSI commands, replaces expensive SCSI cards and cables with ubiquitous ethernet hardware, and permits multiple computers to connect to a single drive array. With iSCSI, your storage location no longer needs to be close to your servers. Since ethernet cables can be up to 100 meters long, you can place a SAN wherever you like.
A key feature of SANs is "thin provisioning," which in effect lets you lie to your server about how much storage is available to it. For example, you can fool each of five servers into thinking that it has 16TB (terabytes) of dedicated storage available when in reality your SAN has a total of just 2TB of actual disk space.
Of course, you’ll still have to add drives to your storage area network as your total data approaches capacity, but thin provisioning helps you avoid wasting space on underutilized servers or creating new partitions each time you add a physical disk.
SANs and virtual machines are a natural pairing. An eight-core server with 32GB of RAM now costs only a few thousand dollars, and that one box can run a handful of virtual machines effortlessly. Pairing a single computer with a storage area network, you can easily create all the services of a larger data center.
To install a SAN, you'll need a storage area network array with at least two drives, one or two additional network cards per server (two is preferable), an ethernet switch (again, a pair is better), and all of the associated cabling. It is technically feasible to use your existing network for your SAN, but in practice you would never want to. Though iSCSI is a reduced-overhead protocol, subjecting your network to both iSCSI and regular ethernet traffic would quickly flood it .
SAN technology lets you store data across multiple servers with a level of efficiency unmatched by traditional options. A SAN can reduce the number of hard disks by 50 percent or more, which saves money in several ways. Fewer drives use less power, obviously, but they also generate less heat, reducing expenses for cooling systems.
There is a learning curve, in addition to additional upfront costs, but the payoff is the need for fewer disks down the road, less administrative overhead, and lower energy costs.
Michael Scalisi is an IT manager based in Alameda, California.