Tiered Cloud Storage
Cloud Society: In the future, all storage will look like this
We all have a tendency to see things from wherever we are standing at the time. This human trait applies directly to technology adoption, leading sometimes to rash assumptions about acceptability (“They’re all going to love this!”). Other times, vendors and service providers come up with new propositions without realising just how profound their impact might be.
For a recent example we need look no further than Amazon’s Glacier launch (LINK: . This new service offers extremely low cost storage – down to 1 cent per gigabyte per month (LINK: http://aws.amazon.com/glacier/pricing/) – with the trade-off that data may take a few hours to retrieve. While certainly not suitable for video streaming, it has a clear role in backup and archiving – the former for risk management, the latter to offload unnecessary data currently cluttering front-line systems.
This reasonably simplistic offering disguises its more profound nature, which may be summarised by saying: Amazon Glacier cannot exist in isolation. The downside of this statement is that the service provider is unlikely to see massive adoption, at least not initially – Amazon will not experience a sudden rush of customers who were already waiting at the door, wondering what to do with all those 50 gigabyte files they only needed to access sporadically.
As an offload mechanism, very low cost storage needs to work in tandem with other, more expensive storage types. This principle was captured by StorageTek engineers a decade ago, when they coined the term ‘Information Lifecycle Management’ or ILM. The central point was that information value has a decay curve – content and data loses value over time, with a resulting impact on how quickly it needs to be accessed. While this ‘long tail’ model may not apply across the board, it is relevant to plenty of data categories.
The result, as its advocates quickly found, was that the key to successful ILM was to ensure data could be moved as efficiently as possible. As data became less relevant it could be moved onto lower cost storage or even archived onto tape. If it became important again, or in the case of primary storage failure, it could be moved back.
Falling storage costs meant that many organisations did not quite deliver on the full ILM vision – put simply, it was easier to replace a storage tier than build an infrastructure based on migration principles; before long, marketing executives were looking for new drums to bang and the term fell into disuse. However the idea remains sound and if the runes are correct, it will no doubt see a resurgence following Amazon’s launch. For the service to succeed, there will be no option.
Indeed, the challenge goes far wider than simple archiving. The current lack of appropriate data migration capabilities could well be the reason why Gartner cites Storage as a Service as languishing (LINK: http://www.cloudpro.co.uk/iaas/cloud-storage/4367/cloud-society-when-did-storage-service-fall-trough) in the trough of disillusion. Meanwhile, the future success of synchronisation services such as Apple’s iCloud, as they grow beyond basic data to a broader range of content types, depends on a comprehensive yet low cost way of storing information online.
What are we likely to see? Almost inevitably, we can expect some kind of tiered model emerging from each provider, with integration between tiers. For example, Amazon’s S3 will need to offer a migration path to and from Glacier. Over time, the two products will become features of a larger whole, in which punters pay for access speeds and resilience features, not just volumes. That’s not a massive insight, more common sense.
Equally obvious is that we are moving towards a model in which capacity is considered to be ‘infinite’ – at least in terms of meeting requirements. Large and small businesses, employees and consumers all stand to benefit from an approach which charges according to need, rather than raw capacity.
Such a model also resolves the issue of locking in with a vendor whose costs quickly become prohibitive beyond a limited number of gigabytes. A service that charges according to what someone is using right now, as opposed to all the content they have ever created, is far more likely to achieve success.
It remains to be seen what such a service, involving automated data migration between different tiers of cloud based storage, becomes called when the marketing wonks and analysts get their hands on it. Whatever its name however, you heard it here first: get the data migration piece right and the rest will fall into place.