Danny Bradbury, Contributor
The enterprise hard drive has a long ancestry in modern IT terms -- it's been well over half a century since IBM invented what became the modern Winchester-style hard disk drive. Hard drive technology is still very much with us and shows no sign of going away, despite the rise of newer technologies such as solid-state drives (SSDs).
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
Nevertheless, the spinning enterprise hard drive faces significant challenges in the next 15 years, and manufacturers will have to work hard to maintain its attractiveness in terms of hard drive capacity, price and performance.
We'll now look at some enterprise hard drive trends that will position the technology for success in the next decade.
Hard drive density over hard drive performance
Hard drives have traditionally been performance-focused devices, but now they have been overtaken in the speed stakes by solid-state drives. This will cause enterprise hard drive technology to be repositioned in the market over the next few years, with manufacturers focused on density of storage rather than performance, said Andrew Reichman, a senior analyst at Forrester Research.
"No one is talking about 20,000 rpm drives anymore," he said. "Today, state-of-the-art is 15,000 rpm because the physics underpinning hard drive technology create a law of diminishing returns that makes it less attractive to increase hard drive speeds."
Disks shrink in size
Although there are very good reasons why faster drives are unattractive -- spinning large pieces of metal very quickly requires more power and more cooling, for example-- that doesn't mean they can't be made smaller.
Increasing areal density is helping manufacturers to move from 3.5-inch to 2.5-inch form factors. A smaller, denser surface area spinning under the read/write head creates a better power/performance ratio for hard drive manufacturers. Hitachi Global Storage Technologies shipped its first enterprise-class 2.5-inch disk drive last year.
TDK, which makes disk heads for many vendors, has released a roadmap in which hard drive capacity will grow in both the 2.5-inch and 3.5-inch formats. In addition, 640 GB platters will begin to come online commercially starting this month in the 3.5-inch format, pushing hard drive capacity up from 500 GB. In the 2.5-inch formats, we can expect to see 375 GB platters gaining traction in the commercial market, upping capacity from the 320 GB platters that were already shifting out of the manufacturer qualification phase.
Storage managers will appreciate the smaller disk formats because it will enable them to fit more drives into a rack and help them to claim much-needed space and power in what is becoming an increasingly constrained physical environment.
As vendors focus on areal density, they need to find new ways to increase the number of bits they can reliably store per square inch. Last year, achievable areal density stood at approximately 400 gigabits per square inch. Sometime in 2011, it's expected to reach 1 terabit per square inch.
Although hard drive vendors that play in such a competitive market are notoriously sensitive about divulging their roadmaps, a look at upstream hard drive technology providers shows how this will be achieved.
For example, Molecular Imprints, which sells equipment that engineers surfaces at nanoscale for the semiconductor and hard drive industries, is promoting bit-patterned media.
Bit patterning seeks to overcome the superparamagnetic effect, which is what happens when very small pieces of magnetic material representing single bits are flipped arbitrarily under normal thermal conditions, thus making the disk less reliable. By arranging the boundaries of these magnetic bits in a pattern, it becomes possible to make them smaller without suffering this effect, which increases the amount of information that can be stored in an inch of disk material.
Heat-assisted magnetic recording
One well-established technology that has been used to increase areal density in the last few years is perpendicular magnetic recording, which uses magnetic bits perpendicularly aligned in the recording material, instead of horizontally aligned.
But, according to Hamish Macarthur, founder of specialist storage analyst company Macarthur Stroud International, the next step change in technology to increase areal density will come in the form of heat-assisted magnetic recording technology. "That's the thing that will take over from perpendicular magnetic recording to push things forward," Macarthur said. "That will maintain the curve of a 30% gain in capacity/performance per year, which has been flattening off a bit."
This concept, which has been talked about for years, involves the application of heat through the write head during the data writing process. Using heat enables manufacturers to use different materials as the magnetic storage media for their drives, allowing them to take advantage of properties that let them manipulate magnetic particles at even smaller scales.
Although vendors have been working on this technique, it's still likely to be commercially applied only after patterned media has been mastered, at which point the two could be used in conjunction to take areal density to tens of terabits per square inch.
From Fibre Channel to SATA
Fibre Channel (FC) is still a relatively expensive technology designed to maximise data throughput. But where businesses require hard drive capacity over raw performance -- such as in tiered storage -- serial ATA (SATA) is becoming increasingly popular in the enterprise space.
Although it lacks many advantages of the new serial-attached SCSI (SAS), SATA is a relatively cheap technology that can be implemented in modular arrays where density is more important than speed. SATA is catching on in tier-2 enterprise environments where fast I/O performance isn't required, although levels of I/O suitable for tier-1 transactional storage can be achieved with large numbers of SATA drives.
The rise of SAS
Fibre Channel's continued success in the high-performance market is far from guaranteed, however. The emergence of SAS represents a significant challenge to that well-established technology. For example, The emergence of 6 Gbps SAS interfaces gives the standard the ability to cannibalise both 4 Gbps and 8 Gbps FC devices.
But we're already starting to see 6 Gbps SAS devices appear. Seagate shipped a 2 TB enterprise-targeted drive with a 6 Gbps SAS interface in February 2010. Such devices will appeal to companies that need a balance of hard drive capacity and performance, such as those dealing with low-latency video streaming and editing applications.
Hybrid hard drives tap solid-state technology
SAS can help vendors balance performance and cost in enterprise hard drives, but it's only a storage interface. Manufacturers still need a way to increase the performance of the disk subsystem.
One solution is to create hybrid hard drives that use conventional hard drive mechanisms for the primary storage function while using a solid-state component to cache data, thus removing the mechanical constraints of the read/write head. Such drives are now appearing on the market, although the cost of the solid-state component significantly increases their price.
According to Macarthur Stroud's Macarthur, the cost of solid-state technology will fall dramatically over the next few years. "Performance is pushing people to use SSD for caching purposes. That makes it more attractive and will increase capacity," he said. "We're looking at the price point of SSD vs. disk. Today, it's way up, but it will come down considerably in the next couple of years." This will give hard drive vendors a flexible tool with which to modify the price/performance ratio of their drives.
As the emphasis for high-density storage switches from tape to hard drive, the encryption challenges previously faced by tape storage vendors are also migrating to disk.
Even though disk drives may not leave the data centre facility as often as tapes do, they still make that journey even if it's a one-way retirement trip. And where companies may have used software for partial disk encryption in the past, they will now be increasingly attracted to the idea of full-disk encryption (FDE) using technology embedded in the disk firmware.
Embedding the algorithm in the disk's control circuitry enables the encryption to operate at the full interface speed of the drive. This represents a significant advantage over conventional software-based full-disk encryption, which incurs a processing overhead. For desktop applications, such an overhead may be acceptable, but for performance-intensive data centre environments it will be less viable, which is why hardware-based disk encryption is more appropriate.
Seagate claimed an industry first in September 2009, when it shipped a self-encrypting drive aimed at enterprise applications. The system was based on a security protocol developed by the Trusted Computing Group (TCG), a security consortium that includes all of the major hard drive vendors. From this we can infer that hardware-based disk encryption is likely to grow over time.
With power consumption a major concern in the data centre, manufacturers are working hard to introduce disk spindown and idling technology into their storage portfolios. Such techniques, where disks spin down or unload their heads when not in use, have so far been restricted to dedicated massive array of idle disk (MAID) subsystems and midrange arrays.
"The challenge for many customers is that they've only been able to do that on certain disks," said Alec Selvon-Bruce, EMEA eco solutions champion at Hitachi Data Systems. Part of the reason for this is that customers have been wary of reliability issues associated with repeatedly spinning disks down and up again.
But vendors have reached the point where mean time between failure on idle disks is good enough to warrant doing it in a high-performance, high-cost environment. "In the future, we will see that technology on high-performance disks. That will come in during 2010 and 2011," Selvon-Bruce predicted.