Conference - Tuesday, Oct 318:30 - 9:15Registration
9:15 - 10:45Mesa Ballroom
Canyon room
Tutorial: Performance Analysis, Tools and Best-Known Methods on Muti-Chip Module Chiplet based High Performance Computing AMD EPYC Zen4 Architecture
10:45 - 11:15Coffee Break
11:15 - 12:45Mesa Ballroom
Canyon room
Tutorial: Zen4
12:45 - 14:00Lunch (provided)
14:00 - 15:3015:30 - 16:00Coffee Break
16:00 - 17:30 |
Conference - Wednesday, Nov 18:15 - 9:00Registration
9:00 - 9:30Cluster 2023 Opening
Mesa Ballroom
9:30 - 10:30Keynote: Bill Magro (Google)
Mesa Ballroom
AI, Cloud, and the Future of HPC.
Chair: Scott Pakin, Los Alamos National Laboratory
10:30 - 11:00Coffee Break
11:00 - 12:30 -- Parallel SessionsDistributed Machine Learning (Session I)
Mesa Ballroom
Chair: Olamide Timothy Tawose, Lincoln University, Pennsylvania
Accelerating Distributed ML Training via Selective Synchronization
PredictDDL: Reusable Workload Performance Prediction for Distributed Deep Learning
Exact Distributed Stochastic Block Partitioning
Resource Management (Session II)
Canyon Room
Chair: Jesper Larsson Träff, TU Wien - faculty of informatics
DEHype: Retrofitting Hypervisors for a Resource-Disaggregated Environment
SciLance: Mitigate Load Imbalance for Parallel Scientific Applications in Cloud Environments
Generalized Collectives for the Exascale Era
12:30 - 14:00Lunch (provided)
14:00 - 15:30 -- Parallel SessionsSoftware Systems for ML (Session III)
Mesa Ballroom
Chair: Jim Brandt, Sandia National Laboratories
FedGuard: Selective Parameter Aggregation for Poisoning Attack Mitigation in Federated Learning
Prophet: Fine-grained Load Balancing for Parallel Training of Large-scale MoE Models
HIOS: Hierarchical Inter-Operator Scheduler for Real-Time Inference of DAG-Structured Deep Learning Models on Multiple GPUs
Storage Systems and Data Management (Session IV)
Canyon Room
Chair: Sarah Neuwirth, Johannes Gutenberg University Mainz, Juelich Supercomputing Centre
FullRepair: Towards Optimal Repair Pipelining in Erasure-Coded Clustered Storage Systems
Performance Characterization of NVMe Flash Devices with Zoned Namespaces (ZNS)
KV-CSD: A Hardware-Accelerated Key-Value Store for Data-Intensive Applications
15:30 - 16:00Coffee Break
16:00 - 17:00 -- Parallel SessionsDisaggregated Architectures (Session V)
Canyon Room
Chair: Hariharan Devarajan, Lawrence Livermore National Laboratory
Rethinking Virtual Machines Live Migration for Memory Disaggregation
Efficient Intra-Rack Resource Disaggregation for HPC Using Co-Packaged DWDM Photonics
ML for Scheduling and Management (Session VI)
Mesa Ballroom
Chair: Haiying Xu, University Corporation for Atmospheric Research
ExplSched: Maximizing Deep Learning Cluster Efficiency for Exploratory Jobs
Hierarchical Resource Partitioning on Modern GPUs: A Reinforcement Learning Approach
18:00 - 22:00Social event: Meow Wolf
Meow Wolf: House of Eternal Return
|
Conference - Thursday, Nov 28:15 - 10:00Registration
10:00 - 10:15Announcements
Mesa Ballroom
10:15 - 10:30Special address: Tim Randles
Mesa Ballroom
25 Years of Cluster Computing at Los Alamos National Laboratory.
Chair: Frank Mueller, North Carolina State University
10:30 - 11:00Coffee Break
11:00 - 12:30 -- Parallel SessionsCommunication (Session VII)
Canyon Room
Chair: Kengo Nakajima, University of Tokyo/RIKEN R-CCS
Communication-Avoiding Recursive Aggregation
HASpMV: Heterogeneity-Aware Sparse Matrix-Vector Multiplication on Modern Asymmetric Multicore Processors
TopoCommit: A Topological Commit Protocol for Cross-Ledger Transactions in Scientific Computing
Workflow and Data Processing (Session VIII)
Mesa Ballroom
Chair: Qing Zheng, Los Alamos National Laboratory
ProvLight: Efficient Workflow Provenance Capture on the Edge-to-Cloud Continuum
Optimizing HPC I/O Performance with Regression Analysis and Ensemble Learning
A Lightweight, Effective Compressibility Estimation Method for Error-bounded Lossy Compression
12:30 - 13:00Lunch (pickup)
13:00 - 14:00Lunch Keynote: Jesús Labarta (BSC)
Mesa Ballroom
Pushing RISC-V into HPC.
Chair: Frank Mueller, North Carolina State University
14:00 - 16:00Best Paper presentations
Mesa Ballroom
Chair: Sunita Chandrasekaran, University of Delaware
A Dynamic Network-Native MPI Partitioned Aggregation Over InfiniBand Verbs
DoW-KV: A DPU-offloaded and Write-optimized Key-Value Store on Disaggregated Persistent Memory
Uniform Algorithms for Reduce-scatter and (most) other Collectives for MPI
JACO: Java Code Layout Optimizer Enabling Continuous Optimization without Pausing Application Services
16:00 - 16:30Coffee Break
16:30 - 16:45Best Paper Awards
Mesa Ballroom
16:45 - 17:45Poster lightning talks
Mesa Ballroom
18:00 - 20:00
Poster reception
Pecos room
|
Conference - Friday, Nov 38:15 - 9:00Registration
9:00 - 9:30Cluster 2024 Presentation
Mesa Ballroom
Chairs: Taisuke Boku, University of Tsukuba
Kengo Nakajima, RIKEN Center for Computational Science 9:30 - 10:30Keynote: Susan Coghlan (ANL)
Mesa Ballroom
Update on the Aurora Supercomputer.
Chair: Antonio J. Peña, Barcelona Supercomputing Center
10:30 - 11:00Coffee Break
11:00 - 12:30 -- Parallel SessionsGPU and FPGA Applications (Session IX)
Mesa Ballroom
Chair: Florina Ciorba, University of Basel
A Finite-Difference Time-Domain (FDTD) solver with linearly scalable performance in an FPGA cluster
GPU Occupancy Prediction of Deep Learning Models Using Graph Neural Network
Reducing Data Motion and Energy Consumption of Geospatial Modeling Applications Using Automated Precision Conversion
MPI & Networking (Session X)
Canyon Room
Chair: George Michelogiannakis, Lawrence Berkeley National Laboratory
SDT: A Low-cost and Topology-reconfigurable Testbed for Network Research
PiP-MColl: Process-in-Process-based Multi-object MPI Collectives
12:30 - 14:00Lunch (provided)
Conference ends |
Conference Room Layout![]() |