venue: TPDS year: 2025 selected: - title: "HARMONIC: Uncertainty-Aware Multi-Objective Optimization for Energy-Efficient HPC Resource Management" authors: - Kyrian Adimora - Hongyang Sun 0001 tldr: "HARMONIC applies uncertainty-aware multi-objective optimization to jointly minimize energy consumption and maximize performance for HPC resource allocation." why_notable: "It directly addresses the growing demand for energy-proportional HPC scheduling with principled probabilistic models, making it relevant to both system designers and green-computing researchers." - title: "MIST: Towards MPI Instant Startup and Termination on Tianhe HPC Systems" authors: - Yiqin Dai - Ruibo Wang - Yong Dong - Min Xie - Juan Chen 0001 - Wenzhe Zhang - Huijun Wu 0001 - Mingtian Shao - Kai Lu 0001 tldr: "MIST reduces MPI job startup and termination latency to near-instant on the Tianhe supercomputer by redesigning the process-management and communication-bootstrap path." why_notable: "Startup overhead is a significant fraction of short-job turnaround time at scale; MIST's results on a top-ranked system provide a concrete reference for HPC runtime developers." - title: "Scheduling With Lightweight Predictions in Power-Constrained HPC Platforms" authors: - Danilo Carastan-Santos - Georges Da Costa - Igor Fontana De Nardin - Millian Poquet - Krzysztof Rzadca - Patricia Stolf - Denis Trystram tldr: "This paper develops a scheduling framework that uses lightweight runtime predictions to respect power caps on HPC systems while minimizing job slowdown." why_notable: "Power capping is now a first-class constraint on modern supercomputers, and this work from leading European HPC scheduling researchers offers practical, deployable algorithms." - title: "PipeMesh: Achieving Memory-Efficient Computation-Communication Overlap for Training Large Language Models" authors: - Fanxin Li - Shixiong Zhao - Yuhao Qing - Jianyu Jiang - Xusheng Chen - Heming Cui tldr: "PipeMesh overlaps pipeline-parallel computation and communication for LLM training while carefully managing memory to avoid out-of-memory failures." why_notable: "Communication-computation overlap is one of the most impactful levers for LLM training efficiency, and PipeMesh's memory-awareness addresses the key practical constraint." - title: "EfficientMoE: Optimizing Mixture-of-Experts Model Training With Adaptive Load Balance" authors: - Yan Zeng - Chengchuang Huang - Yipeng Mei - Lifu Zhang 0004 - Teng Su - Wei Ye - Wenqi Shi - Shengnan Wang tldr: "EfficientMoE introduces an adaptive load-balancing strategy for Mixture-of-Experts training that equalizes expert utilization and reduces communication bottlenecks." why_notable: "MoE models are central to frontier LLM architectures, and load imbalance is their primary training inefficiency; this work provides both analysis and a practical solution." - title: "SSpMM: Efficiently Scalable SpMM Kernels Across Multiple Generations of Tensor Cores" authors: - Zeyu Xue - Mei Wen - Jianchao Yang - Minjin Tang - Zhongdi Luo - Jing Feng - Yang Shi 0008 - Zhaoyun Chen - Junzhong Shen - Johannes Langguth tldr: "SSpMM delivers portable, high-performance sparse-matrix dense-matrix multiplication kernels that scale efficiently across Ampere, Hopper, and future Tensor Core generations." why_notable: "SpMM is a bottleneck in GNN training and scientific computing; cross-generation portability without performance loss is a significant contribution for the GPU computing community." - title: "IceFrog: A Layer-Elastic Scheduling System for Deep Learning Training in GPU Clusters" authors: - Wei Gao 0064 - Zhuoyuan Ouyang - Peng Sun 0006 - Tianwei Zhang 0004 - Yonggang Wen 0001 tldr: "IceFrog dynamically adjusts the number of pipeline stages (layers) assigned to each GPU during training to adapt to cluster heterogeneity and improve utilization." why_notable: "Layer elasticity is a novel dimension of flexibility in distributed DNN training; IceFrog's scheduler provides measurable throughput gains in realistic heterogeneous GPU clusters." - title: "Elastic Relaxation of Concurrent Data Structures" authors: - Kåre von Geijer - Philippas Tsigas tldr: "This paper introduces a formal framework and concrete algorithms for elastic relaxation of concurrent data structures, allowing tunable trade-offs between consistency and throughput." why_notable: "Tsigas's group advances concurrent data-structure theory with a unifying formalism that subsumes many ad-hoc relaxed designs and enables provable guarantees." - title: "Approximation Algorithms for Scheduling With/Without Deadline Constraints Where Rejection Costs are Proportional to Processing Times" authors: - Olivier Beaumont - Rémi Bouzel - Lionel Eyraud-Dubois - Esragul Korkmaz - Laércio Lima Pilla - Alexandre van Kempen tldr: "This paper derives new approximation algorithms with tight ratios for online and offline scheduling problems where rejected jobs incur costs proportional to their processing times." why_notable: "The theoretical results close open gaps in parallel scheduling complexity and are directly applicable to cloud and HPC batch schedulers that must handle job rejection." - title: "EdgeHydra: Fault-Tolerant Edge Data Distribution Based on Erasure Coding" authors: - Qiang He 0001 - Guobiao Zhang - Jiawei Wang 0003 - Ruikun Luo - Xiaohai Dai - Yuchong Hu - Feifei Chen 0001 - Hai Jin 0001 - Yun Yang 0001 tldr: "EdgeHydra uses erasure coding tailored to edge-node failure patterns to provide fault-tolerant data distribution with low redundancy overhead at the network edge." why_notable: "Fault tolerance at the edge is an increasingly critical requirement, and this system's erasure-coding approach significantly outperforms replication in storage efficiency." - title: "Two-Dimensional Balanced Partitioning and Efficient Caching for Distributed Graph Analysis" authors: - Shuai Lin - Rui Wang 0076 - Yongkun Li 0001 - Yinlong Xu 0001 - John C. S. Lui tldr: "This paper proposes a 2D balanced graph partitioning scheme combined with a caching policy that jointly minimizes communication and replication costs in distributed graph systems." why_notable: "Graph partitioning and caching are co-dependent problems rarely treated together; the combined optimization yields substantial performance improvements with strong theoretical backing." - title: "Towards Universal Performance Modeling for Machine Learning Training on Multi-GPU Platforms" authors: - Zhongyi Lin - Ning Sun - Pallab Bhattacharya - Xizhou Feng - Louis Feng - John D. Owens tldr: "This work builds a platform-agnostic performance model for distributed ML training that accurately predicts training throughput across diverse multi-GPU configurations without per-system profiling." why_notable: "A universal modeling framework from Owens's group removes the need for expensive empirical searches when tuning distributed training configurations, benefiting the entire ML systems community."