XaaS — Everything as a Service: Finance
Although one of the advantages usually associated with cloud computing is cost savings, in an interesting article called “Cloudonomics — The Economics of Cloud Computing” published by Rackspace Cloud University, it is pointed out that there are several mechanisms by which the savings are really are generated:
• Reducing the opportunity cost of operating with technology.
• Allowing the change of a model based on capital expenditure (CAPEX) to one of operational expenditure (OPEX), where such costs are incorporated recurrently to the operation, as we do with utilities such as electricity, water, telephone or internet broadband connection; or as when we turn assets into spending through leasing operations.
• Reducing the total cost of ownership (TCO) of technology.
• By enabling organizations to focus on core activities, thus adding more value to business.
The article argues, based on a Gartner study, that most of the costs of operating technology are not related to application execution support, -where there is the greatest value for the organization-, but in the execution of non-core tasks, such as patching operating systems, running backups, maintaining licenses, and more. Cloud computing modifies this balance, as certain technological support roles and expenditure items -which no longer need to be spent- become savings for the IT operation.
At the time of deciding, the costs of all the alternatives should be evaluated comparatively: i) to operate technology at home, ii) to outsource services (in traditional models), iii) or going to a cloud model.
It also recommends considering the decision factors such as initial capital expenditure, initial setup costs, monthly service and labor costs, as well as a cost projection for the following years.
All these factors change in some ways how the traditional business cases of IT projects are elaborated, in which the greatest impact for the calculation of the NPV and IRR, -with respect to expenditures-, were the items of technological investment, with their annual depreciation rates, and whose impact on revenue was relatively stable, unrelated to the potential variation in demand.
In the cloud, it will be easier to calculate the actual computational use of the different processes, and therefore it generates the ability to allocate more accurately the actual costs to the business units for their products and services, which will also mean for them, the challenge to demonstrate their true profitability for the organization. It will be the work of the team in charge of IT financial management to develop and deploy this new budget model.
There are different pricing models in the cloud now, and new ones continually appear. Some of them are fixed, based on pay-per-use, at a subscription price, or elaborating a package of benefits from a catalog of prices (pre-purchased). Other models dynamically establish prices based on services contracted by the user, purchased volumes (even with elastic capacity to extend resources), or some other based on user preferences. There are also models based on a minimum assured amount (and marginal costs per additional units). Finally, there is a new type of “market-dependent” type of price such as those obtained through negotiation, auctions, performance management, very useful for companies that are large consumers of computational capacity.
For this latest market based model, companies like Amazon allow the possibility of bidding the allocation of capacity. For example, using Amazon EC2 Spot Instances (like spot trading in financial instruments), computational resources can be obtained based on the vendor’s availability, but at a better cost than other modalities. Such decisions must be taken based on operational and service factors, but whose financial impact must be incorporated at the decision time, and possibly end up conditioning it.
The additional capacity required by demand related to seasonal batch processes within the company -for example- could be covered by a bid to the supplier, or otherwise, assume the delayed impact on that process -if the cost of such an increase is not favorable or justified.
It is also possible to consider -with a previous feasibility analysis- models like serverless computing (Function as a Service or FaaS), in which the payment is in relation to the functional use of the application, and not variables like the number of servers, in this way granularity in measurement and payment for use is reached (even in multiples of milliseconds), without the “waste” that is generated in other methods of provisioning and billing.
It will be important for the organization to have a clear understanding of the drivers that suit best for them. Pay-per-use may be cheap at first, but-depending on growth-volumes may escalate to be excessive. Alternatives such as fixed cost for agreed periods (monthly or annual) should be evaluated in this scenario.