Storage capacity planning is all about saving money by accurately predicting when you need more storage. Being able to forecast storage needs makes it easier to make an informed trade-off between saving money on hardware and spending money on constant maintenance. But measuring the rate of growth of storage is just a first step; you can also save money by curbing the rate of growth and making sure that storage is properly managed.
Asking department heads how fast their data is growing is unlikely to get you useful information, said Henry Baltazar, storage analyst at The 451 Group. Instead, it's usually a good idea to do an assessment yourself, he said.
If you don’t already use networked storage, that's often a good first step toward using storage more efficiently, Baltazar said. Networked storage lets several servers share the same physical storage, meaning the storage you buys gets added to its storage pool rather than being dedicated to specific resources. Without networked storage, you may have five or six servers with storage that's 80% full, while other servers have as much empty. You have bought enough storage overall, but it's all in the wrong places.
The two technologies for shared storage are network-attached storage (NAS) and storage area network (SAN). SANs are generally more expensive and richer in features, and they come in two flavors, Fibre Channel (FC) and iSCSI. FC SANs are faster but more expensive because they use a separate, dedicated network to connect individual storage devices. iSCSI uses Ethernet, so you can use its existing infrastructure.
Many SANs also allow you to easily add capacity by attaching new storage devices without having to do a data migration, said Kurtis Lindemann, director of technical services at RoundTower Technologies, a storage consultancy. One of the best practices for storage capacity planning is balancing buying too much storage against spending too much on constant upgrades. Investing in a SAN lets you save money down the road, by shifting the equation in the direction of more frequent, smaller upgrades.
Not all data is created equal: Some needs to be accessed frequently, some rarely and some is needlessly copied. One of your best practices for storage capacity planning should be to make sure that the storage you buy is used as efficiently as possible.
A good place to start is tiered storage, in which data is kept on high-, medium- or low-end devices, depending on how quickly you need to access it. For instance, you can store mission-critical data on fast devices that have frequent or even continuous data protection (CDP) backups, while less vital, but still important, data is kept on slightly cheaper devices. Old or archival data can sit on even cheaper and slower hardware, since you won't need to access it as often.
You should also look at data de-duplication, which reduces the amount of capacity you need by eliminating multiple copies of the same file. There are several situations in which you may have copies of the same file, but one of the most common ones is email. If an email with an attachment goes out to 100 people, data de-duplication software finds those 100 copies of the attachment and stores them only once on the server. The software points the attachments to a single copy of the file.
Techniques like data de-duplication and file compression can significantly lower the amount of capacity you need, depending on the kind of files you store and how they're used. For instance, images won't compress as well as text documents. Data de-duplication can effect a storage savings ratio as high as 20:1 if you conduct nightly backups, which create many copies of the same files, Lindemann said. But even with less data, the rate of growth as a percentage will probably stay about the same, Lindemann said. Once you have those techniques in place, you can use your SRM tools and methodologies to hone the new rate.
Another option to consider is thin provisioning. With thin provisioning, a system is tricked into believing it has more capacity than it physically has. This lets you set up large disk volumes -- and thus avoid time-consuming migrations later -- without having to pay for all of that storage up front. The main disadvantage of disk provisioning is that it requires constant maintenance to ensure you have enough capacity; if the system runs out of physical memory, applications can crash and lose data. If you use thin provisioning, it's important to make sure IT understands best practices for storage capacity planning and monitoring so that they can avoid overbooking drive space.