Companies can make better use of data storage

Storage specialist Orca Technologies reports that many companies are spending money on upgraded storage solutions, instead of examining their data requirements and configuring their storage accordingly. Data storage is eating into IT budgets as stored data volumes increase exponentially. The cost of data storage is increased still further with fault tolerance, data mirroring and striping, and the perception that all data, irrespective of application, needs the highest possible levels of availability and security.

"In a typical commercial environment companies are wasting thousands of rand in applying high availability and throughput to non-critical data," says Orca Technologies technical manager Bill Flannigan. "Rather than buying a storage array and storing all data in a single partition on a single Raid configuration, companies need to examine their data requirements, partition their storage environment accordingly, and apply the suitable Raid and availability configuration to each partition."

The crux of the issue is not to pay for features that won`t be used. "Analyse the profile of data stored in a typical corporate environment, and you will find that some data, such as online transaction processing, requires high availability and high bandwidths," Flannigan expands. "Conversely, video data may not require failover or high availability, but it still needs high bandwidth."

The savings lie in using the appropriate drives and Raid levels for the appropriate function. An OLTP environment would use smaller, high-speed drives with Raid 0+1 to produce a partition with high throughputs but low capacity.

"This configuration is expensive because of the 50% overhead needed for Raid 0+1, meaning that for every 4GB of drive, you only have 2GB of capacity," Flannigan explains. "Therefore, you pay a premium for the faster drives and a premium for Raid overhead. You would not apply this across all data types, simply because there is no need and because of the expense."

A simpler data set, such as an accounting report system, is time-critical and also uses low-capacity, high-throughput disks. However, in this instance Raid 5 would be suitable, and as it has an overhead of only 20%, more data can be stored on the same number of drives. User data, such as documents created on a word processor, would use larger-capacity, slower drives that require only a Raid 5 configuration.

"Data access time is slower, but costs can be cut right down by using the slower, larger drives," Flannigan says.

"Data storage costs can get out of hand if a company does not take the time to organise its data types and identify the type of Raid needed for each type. A hybrid solution will maximise data storage and optimise Raid requirements, and free up hundreds of Gigabytes of disk space. There is no point in spending large amounts of money on the latest, fastest drives, if the data residing on them does not need that functionality."