Your team is developing a backend application for an interactive voice response (IVR) system that handles payroll-related calls. The IVR system has these technical requirements:

• Each customer call is tied to a unique IVR session.
• A persistent gRPC connection is maintained between the IVR system and the backend for each session.
• If the connection drops, the IVR system reconnects, introducing minor latency for that call.

Based on current call data:

• Calls last between 1 and 30 minutes.
• Call volume is highest during business hours.
• Significant spikes occur around predictable events (e.g., paydays or major payroll changes).

The goal is to minimize cost, effort, and operational overhead. Which compute environment should host the backend application?