23 September 2009
There are two things I understand about cloud computing.
The first is that it works as an insurance policy. My house might burn down, my computers all get stolen, my hard disks fail simultaneously, and still I will not have lost any of the data I care most about because quietly every night Jungle Disk looks to see what’s new and sends it off to one of Amazon’s servers – now just over 65Gb. I don’t know where, and I don’t care: I am just confident that they are better able to look after data than I am.
The second is that it acts as a universal access service. If my applications and data are in the cloud, I can reach them anywhere and so, if I choose to let you, can you. For me, this blog is at least as much one of those as it is a communication device: it’s a way of being able to find the things I half remember finding interesting long enough ago to have forgotten the other half. It went public somewhat later when I needed the help of google, which is another feature of the cloud – tools and people can interact to get things done in ways which would be more difficult or impossible in less connected ways. As Ars Techica recently put it:
I’m not just doing the same thing that I previously did but in a browser window—I’m actually having a computing experience that I couldn’t previously have sans cloud.
All of that can generate some powerful enthusiasm. Dan Harrison started a recent blog post by saying ‘I love the cloud’ then rhapsodised his way to conclude ‘It’s a wonderful world, and you should embrace its wonderfully warm fluffiness with open arms’, with a string of useful examples in between.
But there is a third thing which I don’t understand about cloud computing: what does it mean for government – or for that matter any large organisation?
Digital Britain is enthusiastic:
Cloud Computing is a model of shared network-delivered services, both public and private, in which the user sees only the service or application, and need not worry about the implementation or infrastructure. The cloud offers the ability to treat IT as a ubiquitous, on-demand service and to flexibly consume as much or as little as is needed. Cloud services are quickly and easily provisioned online and feature granular service catalogues and user-based pricing.
Traditionally, large organisations didn’t insure. If you have a thousand buildings and a normal risk profile, it’s cheaper to pay to rebuild the one that burns down than to pay a thousand insurance premiums. Insurance is, essentially risk pooling plus overhead. If you are big enough to do your own risk pooling, you don’t need to pay for the overhead. In this context, risk pooling means things such as having not just a data centre, but a back up as well, not just having a power supply at each of the data centres, but a back up as well, not just a data connection but… For this purpose, I am not sure that it’s terribly important whether you have your own or rent capacity from a specialised provider. What’s more important in the hunt for the cloud is that there is nothing remotely new about it.
As a bit of an aside, I remember the relish in the voices of two big system veterans as they discussed, not many years ago at all, the challenge of moving one of the government’s Very Big Databases from one data centre to another. The solution? Hire a lorry, put the tapes on it, drive the lorry where it needed to go. Hire another lorry, put another copy of the tapes on it, drive it on a different route to where it needed to go. The only role of clouds in that story is their potential to slow down the lorries.
No doubt there is room for greater scale efficiencies in the government’s use of data centres than has yet been achieved – even governments probably aren’t big enough to pull off Google’s latest trick of having a data centre with no air conditioning that they can simply turn off when the weather gets too warm because they have enough redundancy elsewhere in their network, and there are certainly arguments that government as a whole will do better with more of a shared services approach to the procurement and use of such services. Even then it’s a safe bet that the competitive market for VME hosting doesn’t closely resemble the large scale standard box model favoured by cloud suppliers. As far as I can work out, it is broadly this that is meant by the idea of a g-cloud. To quote from the same section of Digital Britain:
31. The strategy study has established a route-map towards the creation of a G-Cloud, as part of the rationalisation of data centres used by Government and the wider public sector.This would both allow Government to benefit from the core attributes of Cloud Computing e.g. enhanced user experience, flexible pricing, elastic scaling, rapid provisioning, advanced virtualisation while also maintaining the appropriate levels of security, accountability and control required for most Government systems, and lead to substantial savings in costs.
32. The G-Cloud delivery model would also help make other parts of the Government IT marketplace more cost-effective, flexible and competitive. It would support and encourage the adoption of higher levels of standardisation and sharing of IT services across departments. It would allow Government to provide more cost-effectively for peaks and surges in demand for e-Government services; and it would reduce the barriers to entry to the Government marketplace for application and other IT vendors, including SMEs, who would be able to provide services running on standardised, secure infrastructure without having to incur the costs of establishing and accrediting their own infrastructure (in the same way as companies such as 37Signals have used public cloud facilities).
That is, I think, quite an important change in purchasing and contracting terms. Cory Doctorow made the point in a recent Guardian column that
If you’re supplying a service to the public, having a cloud’s worth of on-demand storage and hosting is great news. Many companies, such as Twitter, have found that it’s more cost-effective to buy barrel-loads of storage, bandwidth and computation from distant hosting companies than it would be to buy their own servers and racks at a data-centre.
That sounds like a shift which happened a while back in other industries. The classic example is tyre manufactuers, who stopped selling tyres to airlines quite a few years ago: what they sell now is landings. The same presumably applies to another trick of the cloud, not only do you not need have your own servers, still less your own data centres, but the capacity you do want can be created and disposed of on the fly.
So where does that all leave us? As so often, I think there is risk of confusion because there are different questions (and answers) about the back end and the front end, and it’s often not clear which is being talked about at any given moment.
As far as the back end is concerned, I can see the attractions in changing the nature of the purchasing model in order to make what is bought still less tangible. But once your unit of purchase is in data centres (or rather data centre capacity), I suspect that the impact of that may be less than it would be for smaller organisations.
At the front end the issues are different. Paul Canning has written about the adoption of Google apps by public sector organisations. The focus there still seems to be on cost efficiency: are Google apps cheaper to run at scale than Microsoft Office? (on which I suspect that if they are, and if that starts to get significant traction, the relative prices will change pretty quickly)
But that still leaves the power of doing things in the cloud which are different and better than what can be done in the more traditional model. For end users and for the non-IT transformational power of the latest generation of tools, that is the central question. And on that, I am unsure what the g-cloud really is or what it will offer.
So I am left with the uncertainty I started with: what does cloud computing mean for government?