User Tools

Site Tools



  • May 27 2024
  • All oral presentations will be 25 minutes in duration, which includes a 20-minute presentation followed by a 5-minute question and discussion period.

Keynote Talk (9:00-10:00)

  • Chair: Koji Nakano
  • Speaker: Hiroki Ohtsuji, Fujitsu Ltd.
  • Title: Transitioning High-Performance Computing Systems through Dynamic Scheduling and Resource Allocation
  • Abstract: This keynote talk will delve into the current challenges in High-Performance Computing (HPC) systems, such as the waiting times and low resource utilization derived from batch queue systems, and the underutilization of GPUs due to their fixed allocation. The speaker will then introduce two solutions that address these issues: time-slice execution and the AI computing broker. Time-slice execution efficiently runs applications by eliminating unnecessary waiting times and idle nodes, thereby significantly improving system utilization and user experience. The AI computing broker, on the other hand, enhances GPU utilization by dynamically allocating GPUs only during necessary stages of an application's execution. This talk aims to introduce these new technologies to optimize the efficiency and enhance the usability of high-performance computing systems.

Break (10:00-10:30)

Session 1 (10:30-11:45)

  • Chair: Koji Nakano
  • Application of Network Calculus Models to Heterogeneous Streaming Applications, Clayton Faber and Roger Chamberlain
  • Data-Driven Locality-Aware Batch Scheduling, Maxime Gonthier, Elisabeth Larsson, Loris Marchal, Carl Nettelblad and Samuel Thibault
  • Combining lossy compression with multi-level caching for data staging over network, Rei Aoyagi, Keichi Takahashi, Yoichi Shimomura and Hiroyuki Takizawa

Break (11:45-13:20)

Session 2 (13:20-15:00)

  • Chair: Satoki Tsuji
  • A Scalable Secure Fault Tolerant Aggregation for P2P Federated Learning, Yujiro Yahata, Keisuke Sugiura and Hiroki Matsutani
  • Accelerating BFT Database with Transaction Reconstruction, Aoi Kida and Hideyuki Kawashima
  • Optimizing Aria Concurrency Control Protocol with Early Dependency Resolution, Yusuke Miyazaki, Takashi Hoshino and Hideyuki Kawashima
  • Shared-Memory Parallel Algorithms for Community Detection in Dynamic Graphs Subhajit Sahu, Kishore Kothapalli and Dip Sankar Banerjee

Break (15:00-15:30)

Session 3 (15:30-16:45)

  • Chair: Hiroki Ohtsuji
  • A probabilistic model for asynchronous iterative methods, Pratik Nayak and Hartwig Anzt
  • Introduction to Computational Quantum Chemistry for Computer Scientists, Yasuaki Ito, Satoki Tsuji, Haruto Fujii, Kanta Suzuki, Nobuya Yokogawa, Koji Nakano and Akihiko Kasagi
  • The Logarithmic Random Bidding for the Parallel Roulette Wheel Selection with Precise Probabilities, Koji Nakano
program2024.txt · Last modified: 2024/05/28 06:31 by nakano