
AI Blueprints
By :

In accordance with the AI workflow developed in the previous chapter, we will first identify the problem, goal, and business case of cloud infrastructure planning. This ensures our efforts are not in vain, that is, we are applying AI toward a useful purpose with a measurable payoff.
Cloud computing is commonly used for hosting long-running services such as databases, load-balancing, and "bursty" workloads, such as sudden web traffic. Costs are usually expressed by cloud providers in monthly or yearly terms. However, cloud computing is also useful for one-off batch processing for AI. Many AI techniques require significant preprocessing of large data sets and long training phases. The processing is typically longer than a bursty workload, but the virtual machines are no longer needed when the processing is complete. In some cases, the preprocessing and/or training may be done on multiple machines in parallel.
Naturally, cloud providers...