Search for: All records

In high-performance computing (HPC), modern supercomputers typically provide exclusive computing resources to user applications. Nevertheless, the interconnect network is a shared resource for both inter-node communication and across-node I/O access, among co-running workloads, leading to inevitable network interference. In this study, we develop MFNetSim, a multi-fidelity modeling framework that enables simulation of multi-traffic simultaneously over the interconnect network, including inter-process communication and I/O traffic. By combining different levels of abstraction, MFNetSim can efficiently co-model the communication and I/O traffic occurring on HPC systems equipped with flash-based storage. We conduct simulation studies of hybrid workloads composed of traditional HPC applications and emerging ML applications on a 1,056-node Dragonfly system with various configurations. Our analysis provides various observations regarding how network interference affects communication and I/O traffic. 

Abstract Multi-year marine heatwaves (MHWs) in the Gulf of Alaska (GOA) are major climate events with lasting ecological and economic effects. Though often seen as local Pacific phenomena, our study shows their persistence depends on trans-basin interactions between the North Pacific and North Atlantic. Using observational data and climate model experiments, we find that prolonged MHWs occur as sequential warming episodes triggered by atmospheric wave trains crossing ocean basins. These wave trains alter surface heat flux, initiating MHWs in the GOA and changing North Atlantic sea surface temperatures (SSTs). In turn, Atlantic SST anomalies reinforce wave activity, fueling subsequent MHW episodes in a feedback loop. This mechanism appears in historical events (1949–52, 1962–65, 2013–16, and 2018–22), highlighting MHWs as a trans-basin phenomenon. Our findings link GOA MHWs to trans-basin atmospheric wave dynamics and identify North Atlantic SSTs as a potential predictor of their duration. 

Abstract The 2023/24 El Niño commenced with an exceptionally large warm water volume in the equatorial western Pacific, comparable to the extreme 1997/98 and 2015/16 events, but did not develop into a super El Niño. This study highlights the critical role of contrasting Northern Pacific Meridional Mode (NPMM) conditions in this divergence. Warm NPMM conditions during the 1997/98 and 2015/16 events created a positive zonal sea surface temperature (SST) gradient in the equatorial western-central Pacific and enhanced Madden-Julian Oscillation (MJO) propagation, driving sustained westerly wind bursts (WWBs) and downwelling Kelvin waves that intensified both events. In contrast, the cold NPMM during 2023/24 induced a negative SST gradient and suppressed MJO activity, resulting in weaker WWBs and limited eastward wave activity, preventing the event from reaching super El Niño intensity. A 2,200-year CESM1 pre-industrial simulation corroborates these observational findings, underscoring the importance of NPMM interference in improving El Niño intensity predictions. 

The evolution of mobile networks toward ubiquitous connectivity envisioned by International Mobile Telecommunications-2030 has caused a surge in control plane traffic. A deep understanding of the control plane’s internal characteristics and mechanisms is crucial for delivering optimal services. However, existing measurements often neglect the control plane or treat it as an opaque box, focusing on overall performance instead of its intrinsic characteristics. In this paper, we introduce a 3GPP-compliant control plane evaluation framework and conduct the first in-depth analysis of the characteristics and overheads exhibited by various network functions (NFs) under large-scale connectivity conditions, based on empirical measurements. We selected three core network systems and conducted performance measurements on 500,000 User Equipment during UE registration and PDU session establishment procedures. We reveal the substantial resource demands and limited scalability of the Access and Mobility Management Function (AMF) and the Network Repository Function (NRF). Furthermore, our analysis identifies a significant need for an enhanced state management mechanism. The insights derived from our measurements underscore the immense potential for optimization within the core network. Key optimization pathways include enhancing protocol stack processing, mitigating potential leverage-based attacks, and implementing an integrated state management framework.