Disks rotate and tape moves longitudinally. In both instances the reason is that the read:write head can't detect signals - magnetised areas of the media surface - merely by being placed over them. What the read:write head actually responds to are signal strength changes; from a low strength to a high strength. To do that it needs to look across the surface in the direction of a track and detect signal strength undulations, the edges of the magnetised areas.
With tape the read:write head is fixed and the tape moves underneath it. With disk, the disk spins and the read:write head moves across the platter from track to track. There is also the point that disk spins at a constant speed whereas with tape the tracks run the length of the tape and so, when the physical end of a tape ribbon is reached, the tape stops, reverses direction, starts again and then signal access continues. Disk doesn't have this stop:start characteristic.
No wonder disk access is faster than tape access.
Flash memory is faster than either disk or tape, but nothing moves. How are its signals detected? Every storage element in a flash memory chip is accessible, every bit. It's as if there are as many read:write heads in flash memory as there are stored bits. This helps make flash much more expensive per bit stored than disk. It is so expensive that it simply cannot be used to store the gigabytes of fast access storage needed by servers and PCs.
These three basic templates of storage have persisted for over a decade and look set to continue even longer. There will be minor evolutions but nothing earth-shattering. Disks will spin faster or their areal density will increase. Optical disks employing lasers and holographic storage technology will come, but they still spin and still use a moving read:write head. The basic paradigm is unchanged.
This means that the speed imbalance between storage (disk) access and CPU-memory access will persist and remain unchanged also. Charles Barnes, chairman of Dataslide says, "Hard drive access speeds are the last major bottleneck in common current computer architectures, with up to six orders of magnitude difference between the CPU speeds and the data access response times, and regardless of caching and predictive data management of all kinds, eventually the data has to be obtained from a storage media which operates at millisecond response times, to CPUs operating at nanosecond cycle times." Not even microsecond response times; that would be bad enough, but millisecond response times.
Types of movement
Our storage media are either motionless (flash memory), spinning (disk) or streaming (tape). But these are not the only types of movement we understand. There is vibration. A tuning fork moves by vibrating its prongs. They move sideways back and forth at high speed and so produce a sound.
A butterfly's wings vibrate; a hummingbird's wings vibrate. Even though the vibrating thing has a start stop motion the average speed of the edge of the vibrating item can be extremely high. The trick is to vibrate at high frequency, move very short distances and have a low mass to move. That means there is little momentum to accelerate and brake. Piezo electronics could be used to produce the vibration.
In theory a vibrating storage medium would move under a signal detector and cause it to respond as signal area edges are detected. If we combined the massive number of read:write heads of flash memory, relatively speaking, with a vibrating storage medium we might produce a storage medium with access speed characteristics more like flash memory and capacity costs per bit stored more like disk.
A British company called Dataslide is developing storage technology employing these ideas. There is vastly more to it than this of course. But a working mechanical proof of concept prototype exists and second round funding is taking place. There is interest and contributions are coming from academia; the technical research areas of universities and technical institutions. There is interest and engagement from manufacturers and potential suppliers in the storage industry.
The innovator's dilemma
Clayton Christiansen, a Harvard Business School professor, has coined the idea of the innovator's dilemma. If a successful supplier innovates it is generally to add features to and improve a product, but not to destroy it by developing radically better technology. That comes from disruptive innovations created by others. Shipowners carrying ice from the Arctic Circle to western Europe did not develop the refrigerators that killed their trade. Sustainable innovation cannot withstand the disruptive innovation of upstarts from outside the ranks of established suppliers.
We have seen and are facing sustainable innovation in the flash memory, spinning disk and streaming tape business. But it does not deliver what we IT system users want. That is radically faster access to disk-like quanties of storage at low prices. We want our storage media to keep up with our CPUs and not cost a mint while doing so.
It might just be that vibrating platters and multiple parallel read:write heads in large numbers is the way to do it. We can't expect disk drive manufacturers to generate this technology; that would mean they destroy the foundation of their businesses. Similarly with flash memory suppliers. The high expense of the stuff does their margins no harm at all. We need a hungry and lean company developing the product from scratch, one with no current product baggage.
For lovers of irony we might note that this feature is about shaky technology. But don't knock it. Hummingbirds hover, they hang in mid-air, because of their vibrating wings. The apparently impossible can happen. A violin's shaking strings produce music. A vibrating storage medium could shake up the storage industry, which could be music to us all.
Find your next job with techworld jobs