SIM-SITU: A Framework for the Faithful Simulation of in situ Processing

International audienceThe amount of data generated by numerical simulations in various scientific domains led to a fundamental redesignof how the analysis and visualization of simulation outputs areperformed. The throughput and capacity of storage subsystemshave not evolved as fast as the computing power in extreme-scalesupercomputers, making the classical post-hoc approach highlyinefficient. In situ processing has then emerged as a solution inwhich simulation and data analysis/visualization are intertwinedfor better performance and greater interactivity.Determining the best allocation, i.e., how many resources to allocate to simulation and analysis respectively, mapping, i.e., whereand at which frequency to run the analysis/visualization, and datatransfer mode is a complex task whose performance assessmentis crucial to the efficient execution of in situ processing. However,such a performance evaluation of different strategies usuallyrelies either on directly running them on the targeted executionenvironments, which can rapidly become extremely time- andresource-consuming, or on resorting to simplified models of thecomponents of an in situ application, which can lack of realism. Inboth cases, the validity of the performance evaluation is limited.In this paper, we present SIM-SITU, a framework for thefaithful performance evaluation of in situ processing strategies.We designed SIM-SITU to reflect the typical features of in situprocessing systems. Thanks to its modular design, SIM-SITUhas the necessary flexibility to easily and faithfully evaluatethe behavior and performance of various allocation, mapping,and data transfer strategies. We illustrate the capabilities ofSIM-SITU on a Molecular Dynamics use case. We study theimpact of different strategies on performance and show how userscan leverage SIM-SITU to determine interesting tradeoffs whenadding analysis/visualization components to their application