DynoSim: Simulating the Pareto Frontier

Yongming Ding, Rudy Pei, Hongkuan Zhou, Ryan Olson, Dan Gil, Alec Flowers and Vikram Sharma Mailthody — May 2026

[Yongming Ding](https://developer.nvidia.com/blog/author/yongmingding/), [Rudy Pei](https://developer.nvidia.com/blog/author/rudypei/), [Hongkuan Zhou](https://developer.nvidia.com/blog/author/hongkuanz/), [Ryan Olson](https://developer.nvidia.com/blog/author/rolson/), Dan Gil, [Alec Flowers](https://developer.nvidia.com/blog/author/alecflowers/) and [Vikram Sharma Mailthody](https://developer.nvidia.com/blog/author/vmailthody/) — May 2026

DynoSim Pareto frontier plot showing explored configurations and GPU-verified configurations.

DynoSim is a workload-driven discrete-event simulation of NVIDIA Dynamo: a Dynamo twin for exploring LLM serving behavior before running full deployments. It brings measured engine forward-pass timing, Mocker scheduler cores, Router and Planner behavior, KV cache effects, and workload traces onto one virtual timeline. In our blog post, DynoSim: Simulating the Pareto Frontier, we show how simulation becomes the inner loop for design exploration: sweep broadly, map the throughput-latency Pareto frontier, shortlist the most promising candidates, and verify them on real clusters.