Storage Futures

Change does not come in isolation, it is usually driven by the need to find solutions to problems. I don't think the data storage challenges have fundamentally changed in the past couple of years. They are still explosive data growth; businesses that demand 100 % availability and blinding performance for zero spend; increasing legislative requirements and the problem of getting to grips with complex technologies. The problem with complex systems is that they are risky, and RISK is a four letter word. The challenge then is to simplify systems and so reduce risk, while still providing the level of service that your customers require.

Is RAID redundant?

There has been quite a bit of discussion about this in the last couple of years. The basic problem is that hard disks are getting bigger and if a disk in a RAID array fails it can take a long time to rebuild it. For example, in a RAID5 6+1 configuration using 2TB disks it is necessary to read 12TB of data to recreate a failed disk, while simultaneously recreating missing data from parity for active I/O. While the disk is being re-created, the whole array is at risk from a further failure. RAID6 helps, but may not be a final answer. We will probably see a new approach to data resilience as introduced by the IBM XIV storage subsystem, which does not fit into any standard RAID classification.

Can we expect new technologies to resolve the challenges?

Flash memory usage will continue to grow, but it will not replace hard drives. It's possible to buy a 128GB flash drive now on a small removable storage card but it's expensive, the cost per GB is about 2 orders of magnitude greater that disk. Hard drives will remain on the scene for the foreseeable future. Even laptop and tablet users whose machines just use flash memory will buy external hard drives to get cost effective capacity.

Storage-class memory (SCM) is touted as the next storage generation and should be much faster, cheaper and greener than existing technologies. However SCM is not a technology in itself but is a collection of about 10 different technologies and at this stage, no-one can predict which of these technologies will be the future. These are some of the technologies in the pipeline

Carbon nanotubes could be used to create high density, non-volatile random access memory chips that could replace everything from dynamic RAM to hard drives. OK, we've heard these promises before any it's early days, but one company, Nantero, has built a prototype.
Holographic storage was first proposed in 1963 by Pieter van Heerden at Polaroid and promised very high capacities with fast access rates. 4 decades later still not a reality but it is still in active development. Colossal Storage was developing a rewritable 3-D volume holographic removable disk media but they went out of business and while their technology was bought out, there are no signs that holography is going anywhere soon.
The next innovation in hard drive technology is bit patterned media (BPM), which could achieve terabit per square inch densities or more. To explain the microscopic in terms of the macroscopic, BPM organises its magnetic layer into small mesas, with deep valleys in between. This helps reduce the magnetic interference between each individual bit of data and so allows them to be packed closer together.
HAMR (heat-assisted magnetic recording) technology is potentially a further step up in disk technology and can theoretically support 50 terabits per square inch. It uses a combination of laser technology and magnetic heads, and an iron-platinum disk media. The laser heats the disk as the magnetic head writes data, which permits a much higher and more stable data density.
SPIN BASED STORAGE uses the fact that electrons can orient their spin in two different states, parallel or anti-parallel, usually called spin-up and spin-down. The two different states have different resistances and so can be filtered by an electronic gate and changed with a magnetic field. The technology is currently called Spin-Transfer Torque RAM or STT-RAM and is being investigated by most of the big storage players, including IBM with its 'racetrack' initiative.
The memristor is a device under development by HP, a resistor that stores information even after losing power. It consists of two layers of the semiconductor titanium dioxide; one layer is conductive and the other is an insulator. You can flip the states of the layers by applying a voltage across the device, and that state change remains if the power is removed. The difference between the states is that the conductor contains tiny 'oxygen vacancies' and the insulator doesn't. The claim is that a memristor holds its memory longer that a flash storage device and because it is simpler it is easier and cheaper to make and it can be switched a lot faster, with less energy.

Back to ground level again

Federated is already becoming a new buzzword, and simply means that a number of disparate devices call all be managed through the same interface. The devices can be from different manufacturers, and can be positioned in different parts of the world.
Federated storage essentially adds a new layer of management above storage arrays, so you can increase the size of your storage by adding more arrays, rather than adding capacity to existing arrays, with out the overhead of managing extra arrays.

Another new buzzword in the Storage Networking is 'unified'. Basically, instead of needing several different types of cable to connect devices together, like Fibre and Ethernet, these will be combined together into a single cable, probably running at 10Gb, with the individual 'pipes' virtualised within that cable. This really pays off when you have several virtual devices housed in a cabinet, like VMware or UNIX LPARS. The cabling can potentially be a single cable to the cabinet enclosure.

VENDOR Showcase