A company has hired a technician to build a cloud infrastructure for a massively parallel application. Which of the following should the technician consider doing FIRST?

Adding GPUs to VMs is specific to certain types of parallel applications (e.g., machine learning, graphics rendering) and might not be universally required or the first step for all massively parallel applications.

A massively parallel application often experiences fluctuating workloads. Configuring VMs to scale automatically (horizontally or vertically) ensures that resources are provisioned or de-provisioned dynamically based on demand, optimizing performance and cost efficiency for parallel processing.

Allocating maximum bandwidth might be necessary later but is not the first consideration; efficient resource utilization through scaling is more foundational for parallel applications.

Selecting the most compute cores initially can lead to over-provisioning and higher costs if the application's demand varies; automatic scaling is a more flexible and cost-effective approach.