Operating Systems: An AI & High-Performance Computing Perspective

<p>Follow Meera Iyer, an ML infrastructure engineer at a fictional AI startup, as she diagnoses why the company's shared GPU training cluster keeps producing slow, unstable training runs. Through three real incidents &mdash; a sawtoothing GPU utilization graph, silent out-of-memory kills, and poor GPU sharing across concurrent experiments &mdash; you'll learn processes and threads, CPU scheduling and the noisy-neighbor problem, virtual memory and thrashing, and how GPUs and parallel compute fit into the operating system's picture of managing shared hardware for AI workloads.</p>

Tools: Processes & Threads, CPU Scheduling, Virtual Memory, GPU/Parallel Computing, Concurrency