Noise in the Clouds - Influence of Network Performance Variability on Application Scalability

Explore a conference talk from #SIGMETRICS 2023 that delves into the impact of network performance variability on application scalability in cloud computing environments. Investigate the challenges of deploying HPC workloads on cloud systems compared to on-premise HPC systems, focusing on differences in network architecture and performance. Examine detailed small-scale measurements of latency, bandwidth, and collective communication patterns, as well as simulations of network performance at larger scales. Compare network and OS noise across four popular cloud providers and three on-premise HPC systems, revealing significant impacts on performance and cost at both small and large scales. Learn about the potential limitations of application scaling due to network noise introduced by different network transport and routing protocols in cloud environments.