At first I was tempted to not post the analysis performed by Google’s Vijay Gill which concludes that an on-premise deployment may be cheaper than an Amazon cloud deployment if usage is high and constant.
It’s not because I don’t agree, but because the statement seems quite obvious to anyone coming from with any operational background.
In general, if demand is constant and predictable, it makes sense to apply fixed resources such as in-house servers or full time employees against the problem. If demand is variable or unpredictable, it makes sense to invest more in variable resources such as use-as-you-go cloud resources and seasonal employees.
But ultimately I decided to post Gill’s analysis because it allows me to remind people that most demand models have both a fixed and variable component.
Furthermore, if fixed on-premises and variable cloud deployment models are the resources of the future, doesn’t it make sense to design hybrid applications today that span across both types of resources: your in-house datacenters and the cloud?
Elastic architecture provides you the scalability you need to span; cloud escrow provides you the ability to choose your deployment model. Stay tuned to DivConq as we continue to explore these concepts and the technologies behind them.