Database Reference
In-Depth Information
media reaches the limits imposed by physics, it opens opportunities for alter-
native technology to take over. For example, magnetic disk storage density
was eclipsed by optical storage in the late 1980s, but was able to improve per-
formance at a far faster pace and achieve higher bit-densities than competing
optical solutions. However, as we edge closer to the limits of perpendicular
recording technology (1 terabit/mm
2
), multiple technology options appear to
compete well. At atomic scales, existing approaches to magnetic storage tech-
nology, including perpendicular recording, are clearly not viable. However,
spintronic devices are able to push past the superparamagnetic limit, and
may yet reset the clock again for magnetic technology.
Notice that tape is far below the storage density curve traced by the disk
technology in Figure 1.3. Helical tape technology will not reach the superpara-
magnetic limits to areal storage density until nearly 2020 if densities improve
at historical rates. The cost per bit of tape storage remains far superior and
differentiated from the cost of low-end disk storage in Figure 1.3, so the tape
systems are unlikely to jump to the kinds of storage densities seen in the disk
storage systems because there is currently little economic incentive to do so.
It also indicates that if there is any market pressure on tape storage, the tech-
nology has a lot of margin to improve storage density relative to the physical
limits imposed by the underlying recording media were it to be challenged by
holographic or enhanced optical storage media in the future.
In the area of desktop disk storage, the storage trends have followed a very
consistent slope on the exponential graph except for the nonvolatile solid state
storage, such as FLASH. The consumer electronics market applications for
FLASH storage have created such rapid increases in manufacturing volume,
that the cost per bit has been decreasing far faster than any of the other
technology options. The costs of such devices are now within striking distance
of the upper end of the consumer grade advanced technology attachment
(ATA) disk systems, and may well be a direct competitor to that market
segment in the coming years if this accelerated trend can continue. If FLASH
becomes competitive with mechanical disks, it will dramatically change the
landscape of storage systems—affecting fundamental design decisions for the
parallel file systems described in Chapter 2. However, much work still needs
to be done to characterize the failure modes of nonvolatile storage solutions
to understand how they will behave for large-scale, high-bandwidth scientific
storage subsystems before progress can be made in this area.
1.3 Disk Storage
The very first rotating disk storage device was the IBM 350 RAMAC (Random
Access Memory Accounting system), which was introduced in 1956. It con-
tained 50 platters spinning at 1200 RPMs and had a total storage capacity
