Rubbery Cloud Data Economics

Posted under: From the Blog 23rd of May, 2010

Please enjoy the following rubbery thoughts and numbers comparing our alternatives for storing what we really care about – our data.  No one wants to key it twice. We can put it in the cloud, on a box we bought personally, or on an organisation’s storage.

Amazon and Google are two mainstream cloud providers who probably give us a look at what scale enables, and probably give us two ends of the cloud cost range. GB month is the most common cloud data unit. So let’s use it to compare cloud storage with more traditional alternatives.

Guess what – cloud is about the same cost as buying your own box and much more shareable, and a zero cheaper than corporate storage provided you get personal with your data.

Amazon’s S3 data storage is about $1 per GB month depending on how much you use and how much you use it. Usage drives up the cost quickly, probably reflecting hierarchical storage costs. Amazon has had plenty of experience making money from the cloud so one assumes this is a full cost plus model with no loss leading. It is also probably based on infrastructure architected originally to underpin their main business, ie more commercially solid, risk-averse, older, less innovative and a bit more expensive.

Google on the other hand seems to be our lowball coster, with a history of free initial quotas, and rep for innovating low costs in their data centres. They recently had to pull back some free developer allocations that were unsustainable under demand, suggesting internal lowball or loss leading. A colleague recently described the Google cost structure as based on “optimistically thin provisioning”. Google Apps offers 25 GB per user for $50 per year and make similar commitments to storage qualities, so around 20 cents per GB month. This includes all their apps although most of these come free with a gmail account.

Google probably has greater appetite for risk, innovation, and probably greater scale economies and ability to cross-subsidise from its mainstream business than Amazon.  Its most basic offering, extra storage for gmail and photos, is 2 cents per GB month.
Mediafire, Dropbox and numerous other file transfer services (there are even top 10 lists) seem to circle around about 10 to 20 cents per GB month depending on features. This seems like very competitive terrain with thinning differentiation for the many fighting for share, and is probably close to cost,  maybe under for some loss leading for share.
SSL encryption, slick sharing and integration, larger file sizes and unlimited activity (except by the size of your pipe) seem key to converting tyrekickers into revenue. So personal cloud data comes in around 10 cents per GB month, depending on your preferences.

Personal storage boxes from hardly normal retailers is under $100 per TB device, good for at least a year or two. Once you do what you can to emulate backup and the other cloud service qualities, with two or three copies, software and personal effort, the cost probably approaches $1 GB year or about 10 cents per GB month. And how shareable is your data – fedexing TB drives? You will probably need some even if you use cloud – probably a combination of personal for the more secret, and cloud for the more shareable. Neat that is about the same cost as low end cloud – competition making our choice as hard as possible.

At the medium/high end large organisations typically hold about 100GB of corporate data per employee at a raw cost of around $100 per year. This rubbery number is based on my observations from the higher education sector where a typical uni will spend in 2010 about $1m in capital keeping about a petabyte of central storage online for up to 10,000 staff. Staff who also have another few PB collectively on their PCs. Depends on what staff are doing of course and a small number can drive load – in unis astronomers and physicists can really burn up storage.

Organisations do a fair job emulating cloud service qualities (they would say a better job). But the writing is on the wall – they are screwed on cost. Most CIOs crow about the progress they have made with virtualisation of their infrastructure over the last few years. In the bigger picture this progress should be seen as baby steps, learning how to articulate corporate IT into the cloud.

Many small and medium size businesses, and small business units in large ones, are confronted by the massive price discrepancy between inhouse and cloud IT service provision, and are going cloud. Differences of one or two zeros on one-to-one service comparisons are countered unsuccessfully by rearguard fear, uncertainty, doubt. For my own business I can buy accounting software for about $1000 and then pay for a computer etc to run it on, and a person’s time to run it. Or I can pay $50 a month for software, and all the computing, storage, backup, etc thrown in. So I go cloud and keep a copy of my data so I can sleep at night.

Back to corporate cost. Most internal IT groups are weak at life costing. Once staff, software, maintenance, facilities and capital are added we can add another zero to hence cost about $1 per GB month. Older infrastructure, risk-averse, less innovative platform. More innovative internal IT are looking at how they can build out from here to exploit the cloud, generally using virtualization, federated approaches to inter-organisation sharing, and experimenting with cloud where risk profile allows. Most unis for example have put student email into the cloud using Gmail or Windows Live – small savings but massive service improvements.

So in summary, it seems Amazon targets organizations at a comparable $1 per GB month, aiming to be the high safe corporate road into the cloud. And Google and many others target the personal user and up at around 10 cents per GB month.

As discussed in Cloud Juice Hypothesis, exploiting the cloud shifts the emphasis to personal productivity. The less corporate, the less clumsily secret you can be, the more cloud (knock a zero off your costs, improve services, access innovation) value you can precipitate. But this is long term culture shift. What should organisations do in the meantime, and as catalyst for longer term? As above best organisational practice and Amazon economies are not far apart, so start from where you are.

Build on natural coalitions of organisations with strong data sharing links. Find those coalition members with the most contemporary cloudish infrastructure, latest facilities and closest to biggest pipes. Make selective investments in those platforms, extending them for coalition benefit, as a kind of private cloud. Use the coalition as forum for larger scale experimentation and better articulation from private into the public cloud, and for first cultural steps away from immediate member ownership of infrastructure. As in genetics, hybrid strength.

Or just wave the white flag and pay Amazon, the most expensive must be the best, n’est ce pas?  What do you think?

Add your comment