Costing the cloud: Make the money add up

A basic understanding of the economics driving cloud computing is important, especially given the difficulties in properly costing and comparing the many options open to company IT providers.

An initial look at the economics of cloud computing doesn't make its logic clear. Cloud providers use the same basic components as any other large-scale IT system — much the same servers, storage and networks — but have the additional expenses of running their company and are at the end of internet connectivity that doesn't come for free.

Indeed, if all corporate IT made steady, uninterrupted use of its resources, then cloud computing wouldn't work. But that is almost never the case. Instead, most enterprise computing tasks are 'bursty': they use significant amounts of resources but in short runs, leaving the hardware and software unused in between. Those resources, moreover, have to be specified to cope with peak demand, as far as it can be predicted.

Cloud computing, especially the infrastructure as a service (IaaS) variety pushed by Amazon, VMware et al, works because it only charges for those parts of it that the enterprise uses. Statistically, the cloud providers engineer their resources so that there's always enough to cope with peak demand from individual customers, but on average there's a good match between what's available and what's used overall. That way, the providers average out the use of their infrastructure and can afford to sell it on in an on-demand form.

Resource use

To work out whether a particular task or organisational requirement makes economic sense in the cloud, it's necessary to understand its pattern of how many resources it uses versus how many — and how often — it leaves them idle. Typically, this may be tied to the working day; with two bursts of intense use from morning until lunchtime, and post-lunch to end of play. Adding up to around eight hours out of 24, that's a use:non-use ratio of 1:2; you may also see this referred to as a duty cycle of 33 percent.

A cloud provider will charge you for the amount of computation done plus the persistent storage you need to maintain your server instances during the fallow period. It's quite simple to cost that across a period of time, and then compare the total to the cost of owning your own hardware and software for the same period. Depending on how your company depreciates IT cost, three or five years may be appropriate.

It's important to select instances that are a good match for your own hardware configuration, with CPU and memory properly specified, as those directly affect the utilisation you can make of the cloud service. For non-interactive tasks — ones with a significant amount of computation not tied to direct human involvement — this can drastically change the duty cycle and thus the underlying economics.

Also, it is still, at the time of writing, easier to build a physical server with very large (64GB and upward) amounts of actual memory shared between multiple processors than it is to specify similar instances in the cloud, and for some jobs, having to repartition a task to use multiple smaller pools of memory can have significant performance implications.

Cost factors

Further complications come from factoring those costs that the cloud provider will bear — such as staff, electricity, facilities and so on — but which an organisation must pay for itself if it is self-hosting. Many organisations don't have a good grasp of these figures, and it can have benefits far beyond the IT department if they are quantified and understood.

While there are more alternatives than ever before in IT strategy, there is no alternative to having the basic numbers right and being prepared to use them.

Once you've got all of the above factors understood, it becomes relatively easy to run sets of figures through spreadsheets to identify places where moving to the cloud makes clear economic sense, and where the raw numbers don't add up. Changing duty cycles, expected discount on hardware costs, the cost of up-front money spent on hardware, and other factors will quickly highlight moves worth investigating.

Another benefit of this quantitative approach to analysing cloud economics is the additional leverage it gives you when negotiating deals with suppliers of hardware, software and services. Especially during times like today, when a switch of approach may be permanent and hard-nosed dealing is virtually expected, suppliers should be sensitive to well-argued cost cases and prepared to change their part of the equation to tip the scales in their favour.

While there are more alternatives than ever before in IT strategy, there is no alternative to having the basic numbers right and being prepared to use them.