This article can also be found in the Premium Editorial Download "IT in Europe: Data dedupe: A natural fit with backup."
Download it now to read this article plus other related content.
The modern data centre is a bit like a giant PC. The CPUs reside in one part of the shed, rack upon rack of blade servers providing the processing oomph, while storage sits elsewhere in the building and all are linked by the network/fabric. Sometimes, in fact, the storage devices aren’t even in the building but live off-site in the form of cloud storage solutions.
It wasn’t always like this, of course, for if you go back into the mists of time (say, eight to 10 years ago) even the biggest data centres running open systems had no choice but to buy servers with storage directly attached.
The data centre then was less like a giant PC and more like a shed full of individual PCs. In that regard there is a parallel with the very first factories of the Industrial Revolution, which were simply a gathering together of individual artisans in their workshops.
With the advent of racked server processing and storage area network (SAN) shared storage to support it, CPUs and storage no longer sat in the same box. The giant shed became like a very large PC case; the data centre had evolved, just as the first factories went from being an agglomeration of artisans to buildings that housed machinery-driven end-to-end processes.
Nowadays, manufacturing industry is integrated across a huge geographical area with -- in the most advanced industries -- components scheduled to arrive from multiple suppliers at the production line at exactly the right time to be fitted to a vehicle, for example, pre-ordered by a customer with details specified right down to its colour, trim and optional extras.
The logic of manufacturing economics has propelled industry toward being a massively geographically dispersed operation based on outsourcing and the feeding into the main assembly process of carefully orchestrated inputs.
If the logic of production engineering dictated such an evolution, does the logic of information technology and networks also dictate that the data centre as we know it today should also become just one location -- the data assembly line -- in a geographically dispersed process? Will the advent of the cloud as a location for data storage begin to pull apart the data-centre-as-giant-PC model? Will the giant PC become in effect a rather large thin client?
Well, there’s a certain elegance to the idea -- economy of scale-wise -- that suggests data storage could be carried out by providers in the cloud that deliver data to businesses as required for processing.
A fanciful idea? Not completely. Storage costs a lot, perhaps eating up the biggest chunk of budget in the data centre. Once you have bought your arrays, you have to keep adding disk to them as data growth increases year-on-year. Then you have to power them, cool them and manage them.
The principle of economy of scale suggests that cloud storage solutions could do all these things at less cost than an individual business.
It will happen, of that there’s no doubt. And if you think about it, processing could also be outsourced to the cloud too. The key question, however, is when?
In short, the real potential of the cloud will arrive when network bandwidth improves way beyond the levels we’re used to seeing right now and, possibly more importantly, when we see the advent of transmission protocols that are more efficient than TCP/IP.
As things stand, cloud storage and related services like cloud backup are limited for the foreseeable future. Sure, you can use the cloud to back up to. You can back up with the cloud as a target either via a backup product or using a so-called cloud gateway that creates a virtual network-attached storage (NAS) device that your network sees as a target but actually only exists as a location in the cloud. You can also treat the cloud as an archive target. But, while you can do all that, it’s highly likely you wouldn’t do your first full backup or migrate your archives over the wire as current bandwidth limitations mean it would take an age.
Sure, SMBs can currently work almost completely in the cloud using the likes of Salesforce, Google Docs or Microsoft Office 365, but the enterprise user who requires the speed and guaranteed delivery that come with Fibre Channel is a world away from such SMB-type services. One day, decades in the future most likely, the cloud will be all-pervasive and the suppliers of processing and storage “components” will supply businesses with their data needs for production apps (security/privacy concerns notwithstanding).
Efficient logistics and computerised scheduling enabled the build-to-order orchestration required for modern discrete manufacturing. The availability of high-speed, low-latency, guaranteed delivery networks will allow processing and storage to be independent of business, but for now that’s a long way off.
This was first published in May 2011