Narses: A Scalable Flow-Based Network Simulator
TJ Giuli, Mary Baker
Abstract
Most popular, modern network simulators, such as ns, are targeted towards simulating low-level protocol details. These existing simulators are not intended for simulating large distributed applications with many hosts and many concurrent connections over long periods of simulated time. We introduce a new simulator, Narses, targeted towards large distributed applications. The goal of Narses is to simulate and validate large applications efficiently using network models of varying levels of detail. We introduce several simplifying assumptions that allow our simulator to scale to the needs of large distributed applications while maintaining a reasonable degree of accuracy. Initial results show up to a 45 times speedup while consuming 28% of the memory used by ns. Narses maintains a reasonable degree of accuracy -- within 8% on average.
Create a lesson
Related papers
Scaling Fourier-Based Sparse Matrix Analysis on GPUs
Ruifeng Zhang, Sai Krishna Teja Varma Manthena, Jiajia Li et al.
Efficiently Distributed Federated Learning
Gianluca Mittone, Robert Birke, Marco Aldinucci
Whittle index approach to multi-server scheduling with convex delay costs and impatient customers
Samuli Aalto
PrefixBench-H100: Characterizing Prefix Reuse and Time-to-First-Token in H100 LLM Serving
Omkar Shewale, Deepak Kumar, Divakar Kumar Yadav
Rosetta: Automating First-Principles Performance Modeling Using Multi-Agent LLMs
Karthikeyan Sankaralingam
Optimal Scheduling in Generalized Switch in Heavy Traffic
Runhan Xie, Ziv Scully, Rhonda Righter et al.