Cloud computing provides CPU, memory, network, and storage as resources charged according to use, sometimes down to the second. You can consume them in server-like configurations or as individual services without managing the underlying physical hardware. On-premises computing differs: the organization owns and manages that hardware. section 8.3
The appeal is much like buying electricity rather than running a power plant: draw capacity when needed and pay for what you use. A cloud-native application is designed to scale up and down with demand. That flexibility can help a growing company avoid a large initial investment in servers, but a poor choice can still send costs spiraling. section 8.3 section 1.2.2
Choose based on the workload and the organization's existing investment, not a blanket preference:
- Favor cloud capacity for applications with hourly, daily, or seasonal peaks and troughs, where scaling down matters as much as scaling up. section 8.3
- Examine on-premises hardware for consistently intensive computing, particularly when substantial hardware is already owned. Compare the full costs rather than assuming either option is cheaper. section 8.3 section 8.3.1
- Include security work and data-transfer charges in the decision. Cloud services offer security mechanisms, but using the cloud does not make an application secure by itself; downloading large amounts of data can also incur network charges. section 8.3.1
When the bill rises
Suppose developers create oversized services and leave them running. The bill rises without a corresponding increase in customers. Contrast that with a cloud-native product that scales up as new clients arrive: its higher bill may reflect healthy growth. Investigate the cause before treating every increase as waste. section 13.5
Set spending thresholds for daily alerts, then review running resources at least monthly. Tag services by project or department so their costs have an owner and a purpose. Check whether each resource is still needed, and revise the budget when actual usage gives you better evidence. The practical question is not simply whether cloud is cheaper; it is whether the chosen design delivers the required service at a cost the organization can explain and sustain. section 13.5 section 1.2.2
