While numerous new technologies, from virtualization to converged infrastructures, have promised, but not always delivered, saving in IT operational costs, one radical architecture can tame the IT budget. Providing 50 percent savings in the area with the most cost growth – storage – and help CIOs shift resources from keeping the lights on to investing in innovation, that architecture, writes Wikibon CTO David Floyer in “The Emergence of a New Architecture for Long-term Data Retention”, is a combination of flash and tape in a software-defined environment that he dubs “flape”.

The basic argument he presents is that while flash and tape technologies are advancing in performance and flash is rapidly dropping in cost, disk performance is actually decreasing as data densities increase, and its cost is static. Disk technology is reaching its practical limits in data capacity, and data itself is growing at 50 percent per year. The present architecture in which virtually all data resides on disk is inherently cost inefficient, given that 70 percent of data is only occasionally accessed. And while vendors are investing in advancing both flash and tape technologies, they are putting no money into disk development.

Where data goes to die

 .

Disk is “where data goes to die,” Floyer writes. Disk arrays are growing at 20 percent a year, but given the inherently limited IO available from the technology, the rate at which data can be accessed is only growing at 10 percent, making disk “where data goes to die.” Flash, by comparison, has huge amounts of IO, and that grows linearly as the amount of flash grows.

Flash is more expensive than disk on a per-terabyte basis, but that cost is dropping quickly, driven mainly by high production volumes because of its use in consumer products, from smartphones to tablets and high-end laptops. Soon, Floyer says, the cost of flash will reach that of disk, which is stable.

Tape costs a fifth of disk, is 275,000 times more reliable, has greater longevity, is portable, its areal density is growing faster and vendors are making considerable investment in improving the technology. The Linear Tape File System, invented by IBM in 2008 and released to the open source community, makes tape easy to integrate into the storage infrastructure. And for reading large files, time to read from first byte to last byte is actually faster on tape than disk for large contiguous files, if the tape is already mounted and online in the infrastructure. Disk is still faster than tape (but not flash) in access to first byte, but that advantage is being minimized by new technologies that for instance use multiple heads.

Intelligent tiering

 .

The “flape” architecture keeps the newest, most active data and all metadata on flash, with the remaining 70 percent that is seldom accessed on tape that is part of an active library, which is on the network and mounted, not packed away on a shelf in a vault. With modern technology, Floyer says, this architecture can tame the cost of data growth, cut overall IT infrastructure costs significantly, and provide a dependable quality of service that meets most business needs.

This, he writes, does not mean that disk drives are dead. The flash/disk combination is more efficient in environments handling smaller files – Floyer’s example use case is a medical provider accessing longitudinal patient records.

“Flape” is most efficient where large files are involved – Floyer cites another use case with a media company maintaining digital feature films and other large media files. However, with both flash and tape technology advancing, while disk, where improving efficiency and IO would be difficult and expensive, is not, over time disk is going to fall further behind the flape architecture. Unless something changes, the message is not hard to read.

Like all Wikibon written research, David Floyer’s report on flape, which provides detailed analysis of the points summarized here, is available without charge on the Wikibon Web site. www.wikibon.org IT professionals are invited to register for free membership in the Wikibon community, which allows them to participate in Wikibon research and publish their own questions, comments and research.

feature image: travellingred via photopin cc

photo credit: ViaMoi via photopin cc