Simulate Before You Scale: High-Fidelity LLM Inference Planning Without GPUs

Dipanwita Guhathakurta, Mert Toslali, Srinivasan Parthasarathy +6

Deploying large language model (LLM) inference at scale requires jointly selecting hardware layout, parallelism strategy, scheduling policy, admission control, and request routing…