Overuse of disk storage systems limits efficiencyDate: Oct 16, 2012
Jon Toigo, CEO and managing principal of Toigo Partners International, and chairman of the Data Management Institute, introduces "disk everywhere" dogma in the third tip of his seven-part series. Watch the video above or read the text below to learn why adopting too many disk storage systems can impact the data center.
Please read Toigo's entire video-tip series on data management issues
Examining your data management strategy can have a positive impact on curbing data growth
Capacity management planning is more important than ever for dealing with server virtualization
RAM, hard disk and tape: Which is best for each storage tier?
Value-add software addresses capacity allocation efficiency, but not utilization efficiency
Cloud 'silver bullet' solutions don't always solve cloud storage limitations
Storage decision makers have a lot to consider when dealing with the data storage infrastructure
A huge contributor to the storage infrastruggle (a term I use to explain the challenges IT administrators face) is the impact of dogmatic predilections to use hard disk storage to store all data. "Disk everywhere" dogma is driving organizations to use data storage technology inefficiently and in ways that will impact IT financial performance.
Current analyst research suggests that, by 2011, companies were using approximately half of the 21 exabytes of storage they had fielded to make copies of the files and data stored on the other 50%. Enterprise hard disks and systems built on the technology comprise only one of several data storage methods, and the cost of disk storage systems makes them among the most expensive mechanisms for data storage, especially long-term data storage.
Considerable gains have been made over the years to drive down costs and drive up capacity at the disk drive level. Disk capacity doubles about every 18 months and until recently the price per gigabyte tended to fall by roughly 50% every year. That dynamic should favor hard disk usage as a storage modality. However, disk drive capacity/cost benefits aren't passed to the consumer when drives are cobbled together into disk storage systems. Instead, the cost of disk storage tends to increase year over year by approximately 120% at the array or system level.
Arguably, disk inflation (the practice of using more hard disks) began as a response to performance, rather than capacity, requirements. To obtain greater input/output operations per second (IOPS), vendors discovered that many disk drives could operate in parallel under load. Industry engineers then discovered short stroking (limiting the number of cylinders used on each drive) as a way to eke out a few more thousand IOPS. Ultimately, enterprise performance storage came to be a composite of hundreds or thousands of parallelized short-stroked disks consuming an extraordinary amount of electrical energy to support high-speed operations.
Disk isn't only used for performance storage, but for long-term storage -- especially of files that have ceased to be re-referenced with any sort of frequency. The use of still more high-capacity, lower cost spindles to store what has become archival data makes virtually no sense. In addition to contributing to electrical power consumption, the failure rates of disk drives, the impact of silent corruption as evidenced in bit error ratios, and the availability of more cost-effective alternatives (like tape and newer tape NAS offerings) make capacity disk arrays a questionable fit for archive applications.
From the standpoint of disk-to-disk (D2D) replication as a hedge against data loss, the strategy has merit, but only when it's appropriate to the recovery requirements of a business process and usually only with the additional safeguard of a tape backup. Since comparatively few applications require instantaneous restore if an interruption event occurs, only these applications provide the cost justification for expensive D2D backup in a fact-based world.
Finally, it's important to note that, in the absence of a common management scheme that works across all array products from all vendors, the disk-everywhere meme is a road to higher labor costs for storage infrastructure management, which contributes to the overall cost model of enterprise storage.