By Jon Collins

 

 A couple of years ago, I was hosting a session at Storage Expo. One of the presenters was from a major financial institution, and he was extolling the virtues of what was then a brand new storage technology, called Massive Array of Idle Disks, or MAID. Truth be told, there wasn’t all that new about the technology – it still involved racks of hard drives, big boxes with flashing lights, and all that. Where it differed was in one simple but quite fundamental way: that the hard drives, when not being used, could be turned off.

 

I had heard little about MAID between then and now, and indeed, had I given it any consideration I would probably have written it off as yet another technological also-ran (like RLX, the server blade manufacturer whose blades could also be switched off when idle - the company is now part of HP). The premise seemed to be based around the word “massive” – i.e. you’d need lots of disks and a great deal of data, most of which you’d never use, before you’d start to see the benefits. And while companies might be able to identify specific applications to take advantage of such a configuration, the majority would probably be happy with what was more widely available.

 

Much can change in two years, however. For a start, the continued erosion of disk costs and higher capacities, together with the more general availability of Serial ATA drives (more bytes per buck than Fibre Channel) has caused many companies to think again about how they tier their storage – in particular, whether application owners can be convinced to migrate over to the cheaper disks. Second (and again due to wider availability of cheap disk), there has been the growing uptake of virtual tape – that is, disks that appear as tapes to backup and archiving applications. Third we have the wondrous power of the compliance wave (what a boost to the storage industry that was), requiring data items of various types to be stored for longer, and rendered more readily accessible.

 

Last but not least, and hot on the heels of Messrs Sarbanes and Oxley, we have the latest green fad. Don’t get me wrong, I’m all for anything that helps keep climate change at bay, and often, power efficient solutions will be good for company economics as well. In marketing terms however, there is a green-painted wagon rolling down the hill, and its worth jumping aboard right up to the point that the wheels come off and attention gets turned elsewhere.

 

From a MAID perspective, all of these factors conspire to turn around what was once a technology looking for a killer application. The need to maintain large pools of data, much of it will never be accessed but equally, which does need to be kept reachable in case of demand, coupled with continued application consolidation and a need to keep power costs as low as possible, results in MAID becoming quite an attractive proposition. Virtual tape is a clear opportunity, but so is the more general requirement for longer term, accessible storage of content. There may still be application-specific uses, but there is plenty to suggest that MAID is more than a niche capability.

 

Indeed, there are other stated benefits, such as increased reliability (a counter to the unreliability of disks due to their moving parts, is to ensure the parts move only when necessary), the ability to pack the disks more densely (less power means less cooling) and so on. While the mathematics in favour, or against these benefits are still to be fully ironed out – for example, see the recent study by Google – they should at least be investigated – particularly by storage managers who fear that turning disks off could increase the risk of disk failure. As it happens, it is power cycling that Google cites as having the potential to increase failures, which should be against the principles of a well-designed MAID system anyway.

 

MAID shouldn’t be seen as the panacea to all storage ills, and it is still a work in progress, particularly when it comes to assuring that the disks are used as efficiently as possible. Companies that deliver MAID such as COPAN Systems may have the hardware and base software in place, but they still need to address areas such as virtualisation, policy-based data movement and intelligent indexing. Most likely, these areas will be covered by the larger vendors (EMC, HDS) and through collaboration with specific software suppliers (Datacore, Falconstor, Njini, FAST Search and so on). Neither is tape, or even optical disk dead – both of these offer significantly better energy profiles than racks of constantly spinning disks, though with the management overheads of media handling. All the same, however, just as all of these technologies have their place in the storage architecture, so should MAID.