Revati: Transparent GPU-Free Time-Warp Emulation for LLM Serving

TL;DR

Revati is a GPU-free time-warp emulator that enables fast, accurate performance modeling of large language model serving systems by virtualizing GPU execution and allowing rapid simulation of different configurations.

Contribution

Revati introduces a novel time-warp emulation approach that directly executes real serving system code, virtualizes GPU management, and performs fast-forwarded simulation without requiring physical GPUs.

Findings

Achieves less than 5% prediction error across models and configurations

Runs 5-17x faster than real GPU execution

Reduces testing costs and time for LLM deployment

Abstract

Deploying LLMs efficiently requires testing hundreds of serving configurations, but evaluating each one on a GPU cluster takes hours and costs thousands of dollars. Discrete-event simulators are faster and cheaper, but they require re-implementing the serving system's control logic -- a burden that compounds as frameworks evolve. We present Revati, a time-warp emulator that enables performance modeling by directly executing real serving system code at simulation-like speed. The system intercepts CUDA API calls to virtualize device management, allowing serving frameworks to run without physical GPUs. Instead of executing GPU kernels, it performs time jumps -- fast-forwarding virtual time by predicted kernel durations. We propose a coordination protocol that synchronizes these jumps across distributed processes while preserving causality. On vLLM and SGLang, Revati achieves less than 5% prediction error across multiple models and parallelism configurations, while running 5-17x faster than real GPU execution.