Data Storage Topics

Ezra Strong
CS 441


Introduction


As it stands, if you were to go to the store to buy some form of data storage, you'd be presented with two choices, one being a traditonal hard drive, made up of platters of material, which are then used to magnetically store your bits. The information is stored by moving a "head" over the areas of disk that you wish to deal with. Part of what characterizes these devices, in general, is the word magnetic -- these are magnetic storage devices.

Your other choice is often called "flash" memory. More correctly, it is called solid-state memory. This sort of memory is controlled by sending currents into the chip -- there are no moving heads or parts on the mechanical level. These chips use semiconductors to store their signals -- this is an electric method, not magnetic. I'll talk about another side-effect of this momentarily.

Each of these has a set of trade-offs associated with it. The hard disk has mechanical and moving parts that can fail, the solid-state memory has a limited number of read-write cycles. The solid-state memory is faster, but each unit of space is more expensive in terms of money. This is due to the way they store the actual data -- retrieving a signal from a transistor is a relatively fast process -- measured usually in tens of microseconds. We measure the access time for our magnetic hard drives in milliseconds. This is 100 times slower than our flash memory.

These aren't the only way of storing data. One fairly popular way is with optical drives -- the things that read and write to CDs and DVDs. CD's store their information in what appear to be "bumps" along small tracks in the disc. To read from it, a laser scans over bumps -- the reflection from the laser passing through those areas of data storage can be decoded to reveal the data hidden within. This form of data is characterized by its optical systems for reading and writing. Optical media is also typically removable -- discs for moving around rather than an actual one-piece hardware device of media and read/write functionality all together.

For a simple and graphic overview of any of these existing storage methods, you might want to check out www.howstuffworks.com.

There are, however, new ideas in the field of memory -- new ways of storing our data on a variety of media. We would ultimately prefer that the media we are storing it on overcomes the failures of current medias, of course. Research efforts have been both pursuing ways of pushing back the barriers and going outside our conventional methods altogether. Hopefully, these can shine some light on a couple of the things people have come out with!

Perpendicular Magnetic Recording


As mentioned, our traditional hard drives are forms of magnetic storage. With normal hard drives, the bits on the platters are stored horizontally -- stacked end-to-end. In the past, one way researchers sought to increase the capacity of hard drives like this was to shrink down the individual grains data was stored on. Unfortunatly, this can only go so far -- not only engineering problems, but there's also a point known as superparamagnetism, that depends on the density of the grains, and the coercivity of the material's magnetic field. When this point is reached, the local temperature is enough to cause bits to randomly reverse direction, effectively scrambling your data. As you can imagine, this might be a slight problem. This limit is predicted to now be around 100-200 GB/in^2. Even this is predicted as after reducing the track width per bit length (known as the bit aspect ratio) to around 4:1 -- typical products around the millenium were 20:1. While this may! seem to be a large number, it was seen to be a nasty limit, a hurdle to be cleared. What researchers turned out was surprisingly simple: turn the bits "upright". Instead of stacking the bits end-to-end, longitudinally, stack them upright, so that they're parallel to each other and perpendicular to the disk. This is known as perpendicular recording. It was, somewhat surprisingly, originally proposed as part of an IBM magnetic drive research effort in the mid-50s. Technical and cost considerations at the time forced the project to revert back to longitudinal recording, and perpendicular recording was lost for a time. Additional information about the IBM effort can be located at http://www.magneticdiskheritagecenter.org/MDHC/NAPMRC_HOAGLAND.pdf. In the 70s, a new researcher picked it up -- Shun-ichi Iwasaki. His continuing research into the matter was theoretical -- perpendicular recording drives didn't hit the market until 2005.

Perpendicular bit recording drives not only use up less horizontal space per bit, allowing more to fit in the same area, but also features a layer beneath the bits, using a material that is considered "soft" in magnetic terms. This, in turn, produces helpful effects in terms of keeping the data stable. The orientation of the bits also proves to be more stable as a general thing, resulting from the polarity of the bits helping to hold them. As part of promoting their perpendicular drives, Hitachi produce a small flash animation video called Get Perpendicular, viewable at http://www.hitachigst.com/hdd/research/recording_head/pr/PerpendicularAnimation.html.

Recording the bits like this is not, overall though, a heroic achievement. This simply allows for greater expansion. This does not rid us of the limit altogether. It allows far larger capacities, but they still are constrained, and still are hard drives as we know them, more or less.

Prior to the development of perpendicular drives, the largest commercial drives were around the area of tri-platter affairs for 350-400 GB. Since the introduction of perpendicular recording, you can now purchase 1.5 TB drives for under $200. Smaller drives are also gaining capacity -- laptop drives, and storage for things like MP3 players. Again, we have to consider though, this is a temporary measure, to bide while other research avenues are pursued. Predictions are for roughly ten times the capacity from advanced perpendicular magnetic recording drives, when they become fully developed.

Nanowires!


One of the new techniques that has been being toyed with is what is known as nanowire technology. Nanowires are a form of solid state storage. They are also what is known as non-volatile memory -- this is different from the memory used is things like PDAs, in that it does not need current flowing through it to preserve memory. There has been more than one way of using nanowires for storage proposed and a variety of tests have been conducted with slightly different ideas and goals. First, though, I should cover, what is a nanowire?

The name gives away much of it. Nanowires are incredibly tiny wires -- they can be constructed from a variety of materials, depending on what you are trying to achieve. One thing you can build out of them is semi-conducting transistors -- like we already use to store data! Except, now, it's being built on a nano-scale, instead of microns.

The methods used in assembly are also a part of what makes them special. Several tested techniques include self-assembled molecules or ferroelectric thin films, all of which supposedly require less resources and overhead than other things like... flash memory. Currently, though, these are imperfect test procedures, and as such, there are some problems in the finished product. Other tests that have been done have been attempts at melding current manufactoring techniques with nanowires. One such method -- implementing nanowires with oxide/nitrate/oxide layers -- have been successfully tested: see stacks.iop.org/Nano/18/235204.

The resultant product had a number of beneficial traits, including the ability to retain a charge -- its memory of data -- with little leakage, a large window for memory states, and a large ratio between the input voltages. The complete results can be found in the iop paper linket above. Nanowires are also noted because of their single-crystaline structure, leaving less room for defect.

One thing that helps tremendously for costs in nanowire manufacturing is the fact they are generated via a bottom-up methodology -- that is, the product is envisioned and worked on from the small pieces first, and then put together to make a larger piece, and so on up the ladder. This can be done because of the apparent tendencies of small things like the nanowires to join together. A particularly interesting bit with regard to nanowire research is the tests done by some researchers at the University of Pennsylvania -- they created a nanowire, similar to coaxial cable, in that there was both a core and a shell section -- these nanowires exhibited THREE distinct states. Trinary memory. In one sense, this is a step forward beyond binary, giving up more room in a single "bit". On the other hand, everything right now is in binary -- transitioning may not be easy or cheap.

Holographic Discs


As a personal interest of mine, the final data storage technique I'm going to present is holographic memory. Perpendicular recording is a magnetic storage method, similar to current hard drives. Nanowires are solid-state, storing signals as electronic impulses. Holography stores data via optical patterns, similar to CD and DVDs. They are also similar in that holography is used as removable media, for backups and the like.

To record a holograph -- your data, for example -- you start with a laser, that you split into a reference beam and your carrier beam. The carrier beam shines through and "picks up" your data, essentially carrying it onwards to a surface that can take it. The "picking up" involves passing through a special device, known as a spatial light modulator -- despite how much this may sound like something from science fiction or cartoons, it is essentially just translating your data into light patterns for the beam to store for the future. In the case of holographic drives, the surface you'll be recording to is your disc -- some sort of photosensitive layer of the disc. By shining the two beams back together, you can create an interference pattern on the disc. To read the interference pattern, you shine the reference laser alone at the disc, and then translate what it gets back with proper sensors -- CMOS.


So, we shine lasers at a disc to make it store data. Expensive lasers, and expensive discs, for now. What makes this so special? For one, there's how the holographs can store data. Unlike CDs or DVDs, where data bits are stored side-by-side in tracks, holographic images can be stored overlapping. Holograms can also be projected through nearly the full volume of the media, rather than just a surface recording layer. These combined factors mean that the media should be able to hold very large amounts of data. The retrieval of the data also differs to a noticable extent -- the data from a holographic disc is pulled in "pages", rather than the bit-by-bit way other drives work. Because of this, the bandwidth of a holographic drive should be decently large -- on the order of a gigabye a second. That is, unfortunately, for now, theoretical. InPhase, one of the pioneer companies working on holographic technology, reports far slower estimates with their current drives, o! n the order of 160 MB/s. The size for a single disc is starting around 300 GB per.

The discs are supposedly highly resilient as far as medias go, once the data is encoded. Scratchs, cold, heat, are all said to be relatively harmless -- the advertised lifespan for a disc is a whopping fifty years in terms of lasting time. As reading and writing are done by laser, the systems offer less mantainence required than tape drives.

Right now, such devices are at the industrial and company levels -- high purchase costs and a lack of practical "time will tell" testing have limited the spread of holographic media. Fine-tuning still needs to be done on the technology, as well as a large decrease in production and development costs, before such things hit the commercial market. As with developing technologies, there's also the risk of the product outgrowing its past. With what appears to be a long-term reliablity solution to data storage, you probably don't want to be worrying about whether the future technologies and impovements will support your old media. Whether or not this media ends up becoming a viable solution depends on the market, the researchers, and the marketing. Perhaps we'll see in ten years... if we survive that long.

Additional Reading


Perpendicular Magnetic Recording


Iwasaki
Hitachi paper
IBM History

Nanowires


Core-shell Nanowire(Trinary state) Logic from Nanowires
ONO Chips from Nanowire

Holographic Technology


IBM Research Journal article
Extensive IEEE Paper on Holographic Storage Solutions
JPL/NASA Paper on Holographic Storage Comparison of Bit and Page-based Holography