EdgeCampus runs on three layers: students' phones, a GPU edge cluster in each building, and a public cloud region. Consider three workloads:
- W1 — CampusAR frame recognition: 30 frames per second per student, each needs ~40 ms on a phone CPU or ~5 ms on an edge GPU; the student needs an answer within 100 ms.
- W2 — lecture-video transcription: 90 minutes of video per lecture, needed by the next morning.
- W3 — faculty scientific workflow: 500 tasks with dependencies (a DAG), some tasks run much faster on GPUs.
For each workload:
- Choose the layer(s) where it should run and justify with latency, capacity, energy and cost.
- Name the resource manager that decides (phone OS, edge scheduler, cloud autoscaler…) and the decision it makes.
- Name the metric that tells whether the decision was good (response time, deadline met, makespan, cost, energy…).
- Name one course chapter where you will learn to make that decision better.
Finish with a paragraph: Which of the three problems looks most like a research problem, and why?