Giving Storage Its Due

Don't take data storage for granted; follow these tips to improve your storage repository design.

Greg Schulz, author of several books on storage, is founder of the StorageIO Group, an IT industry analyst consultancy whose blog can be found at storageioblog.com and at twitter.com/storageio.

IT shops always need timely access to stored documents and more space for storing them, but data storage seems to be an issue that often gets less attention than it deserves.

Not all data is the same in terms of how frequently users access it and how it is retained, but it’s typically treated the same. While the cost per unit of storage is decreasing, the amount of storage that can be managed per person is not scaling at a proportional rate, resulting in a storage management efficiency gap.

Here are three tips to improve the design of a storage repository:

1) Reduce the footprint

Options for reducing a storage system’s data footprint include deduplication, real-time compression and thin provisioning.

Deduplication, in which redundant data is eliminated, is a good fit for inactive data, archives or backup where there is a high degree of commonality or repeatable data and where processing time can be traded for a larger data reduction ratio.

Real-time compression is a good fit when time and performance or data transfer rates are important for systems that still need to save some space.

In thin provisioning, storage space is allocated on an as-needed basis rather than being allocated up front. In a traditional system, if 10 users each receive 1 terabyte of storage but use only half of that, then 5T of storage space goes unused. By allocating space as needed, thin provisioning provides the storage users need without wasting space. Of course, as usage increases, additional storage will be needed.

2) Determine the best RAID level for your needs

Different levels of RAID (Redundant Array of Independent Disks) provide a balance between performance and protection. The choice of RAID level can have a significant impact on the performance and optimization of a storage system.

RAID 0 offers speedy performance and availability but no fault tolerance, so any disk failure destroys the array.
RAID 1 can provide fast read performance but slower write performance. Because data is also copied to a mirror disk, this level provides fault tolerance.
RAID 5 provides good performance and fault tolerance, although a disk failure will diminish performance.
RAID 6 offers greater fault tolerance, allowing the array to function if two disks fail.

3) Find the right tiered storage device

Different types of tiered storage devices offer varied benefits. Agencies should determine which types best meet their storage needs.

For example, ultrafast solid-state drives (SSD) offer speedy performance at a price. If you have the need for speed, consider using SSDs based on the whether their speed (measured in input/output operations per second) merits the increased cost. Fast 15,500-RPM SAS and Fibre Channel hard disk drives (HDD) provide a mix of performance and storage capacity and generally cost less than SSDs. With SAS and SATA HDDs, users get higher capacity at a lower cost, but they sacrifice some speed.

Some applications may need high performance, while others require lots of storage capacity. Depending on its size, your storage system may have a mix of SSDs and HDDs.

Another option for smaller systems or workstations is hybrid HDDs that combine the best of worlds: a small SSD and a high-capacity HDD in a single, integrated device.