A company's online retail application experiences erratic and unpredictable traffic, leading to challenges with over-provisioned servers during quieter periods, in turn inflating operational costs. The application currently runs on a constant number of compute instances which are occasionally underutilized. The architect must devise a scaling methodology that aligns costs with traffic levels, ensuring the system scales down during low traffic and scales up to meet high demand without service interruption. What recommendation should the architect make?
Recommend purchasing additional Reserved Instances to cover the maximum anticipated traffic for maintaining constant service levels at all times.
Suggest transitioning to an entirely on-demand serverless model to eliminate the need for managing server capacity.
Propose provisioning a number of always-on instances that correspond to the highest observed traffic levels, turning off auto scaling.
Advise utilizing a dynamic scaling setup with a base layer of fixed instances and supplementary elastic instances for demand spikes.