Your question is Storage Trade-offs for ML. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
What are the trade-offs between various storage solutions for high-throughput machine learning workloads?