Sunday, 11 March 2012

Costs in a Cloud-Computing IT

I’m getting deeper into the study of deploying a cloud-computing solution in my organisation, and to do, I’ve learned some important aspects about financial issues.
 
It seems clear that going “cloud” means that your CAPEX (Capital Expenditure, that is, a one-off investment made for example to buy servers) will be reduced drastically, or even, why not, minimize to zero. Since infrastructure (IaaS or Infrastructure as a Service) is provided externally to my company, I don´t have the need to buy it.
 
But… what’s not so clear is the way that OPEX (Operational Expenditure, that is, monthly cost of operating your business) takes in the financial ladder… though it seems to depend mainly on the use of resources my applications do when running and the length in time of that use. Said that, it means that before going “cloud”, you must at least consider taking these steps:
 
-         Study thoroughly your apps requirements and resources consumption
-         Analyse usage of resources on a monthly / yearly basis (depending on the type of operations your company does)
-         Get as many metrics as you can of the behaviour of your apps
-         Analyse your NPV (Net Present Value) to study your present cash-flow versus the one you’ll have if gone “cloud”
-         And most important of all, never go “cloud” in a big-bang approach, just choose which apps are ready to be deployed in the new model
 
Why do I insist in metrics and consumption? Well, since my financial model will turn completely from a CAPEX model to an OPEX one, I do need to adjust / trim my app performances until I can get the best value of going “cloud”. What do I mean? I mean, if I have apps in my organisation which consume lots and lots of resources but their executions are just planned once a month (as it’s the case in the banking sector), that has to be considered when planning their migration to the cloud, since I must include in the agreement with the provider of my IaaS service that I’ll need resources but just for a day in a monthly basis, and that will mean fewer costs in my OPEX side of the balance sheet.
 
As a real example out of IT, let’s consider we need a car and we are comparing if we should rent it or just buy it. The quickest answer is, well, it depends on the use you’ll make of it. If you need a car just a few days a month, it will no doubt be cheaper to rent it (cheaper CAPEX, if any, and expensive OPEX) but if you use it in a daily basis, you’d better buy one (expensive CAPEX but cheaper OPEX).
 
So, cloud-computing mainly involves taking the path from a CAPEX model, the one we are involved and used to now, to an OPEX model.
 

No comments: