Database Reference
In-Depth Information
of small files when they are migrated to tape to address part of this problem.
However, such systems will need to be refactored to better handle small files
by providing policies that optimize aggregation based on the access patterns
to the tape.
Every few years, there have been white papers questioning the long-term
viability of tapes and the likelihood of their being replaced by spinning disk
storage. Despite this, the cost performance and power performance of tapes
continue to maintain order of magnitude benefit over disks and even NVRAM
devices. Due to their enormous surface area and the ecient storage layout
offered by helical scan heads, tapes maintain high storage density despite being
far below leading-edge areal densities offered by leading-edge magnetic storage
technology, and thus provide very little pressure on vendors to push the limits
on the technology. Were a competing technology to emerge that put pressure
on the tape market, the tape technology vendors have significant headroom
to improve densities and price performance. However, such a competitor has
yet to emerge, and current power, density, and storage trends for disk make
it unlikely to be the likely successor to tape.
It is not clear that there is a viable competitor to the tape market for the
lowest tiers of storage in mass storage systems primarily due to the cost and
power savings that tape continues to provide. However, a potential future com-
petitor to the tape market, namely holographic storage, continues to remain
on the horizon. Holographic storage has been “a year away from production”
for at least two decades, with issues that continue to challenge its ability to
achieve commercial production. Holographic storage is much like optical stor-
age in that it will likely have its niche applications and uses within the storage
industry when it arrives, but will take a while to develop characteristics or fea-
tures that would make it competitive with the demands that high-performance
computing centers require of disk or tape systems. Likewise, other passive
technologies such as direct molecular manipulation (IBM's Millipede) are un-
likely to compete with tape on the basis of cost, density, power, or streaming
performance alone—but they do offer higher performance for random accesses,
which would be more appropriate for storage of large numbers of small files.
In summary, there exist many new and exciting emerging storage technolo-
gies with interesting and unusual storage characteristics. In the near term,
none of these are likely to disrupt the current high-performance computing
industry trends of primarily using disk and tape to store data. As the petascale
computing age begins, the storage industry is likely to see another significant
increase in the amount of data stored and retrieved from these larger and
more capable systems. The leading challenges to disk storage are power and
reliability or data integrity. However, tape will be challenged with increases
in access times and the amount of data at risk as the size of a single cartridge
increases. These are problems that the storage industry is working to solve,
but at some point or scale will cease to be reasonable for any storage device
to handle. This demands solutions, potentially by data management software
to improve on the reliability and access to ever-increasing amounts of data.
Search WWH ::




Custom Search