A5012588973- Distributed and Parallel Computing Systems
- Advanced Data Storage Technologies
- Cloud Computing and Resource Management
- Scientific Computing and Data Management
- Parallel Computing and Optimization Techniques
- Particle Detector Development and Performance
ETH Zurich
2018-2024
Swisscom (Switzerland)
2014-2024
CSCS - Swiss National Supercomputing Centre
2014-2023
Supercomputers have been driving innovations for performance and scaling benefiting several scientific applications the past few decades. Yet their ecosystems remain virtually unchanged when it comes to integrating distributed data-driven workflows, primarily due rather rigid access methods restricted configuration management options. X-as-a-Service model of cloud has introduced, among other features, a developer-centric DevOps approach empowering developers infrastructure, platform software...
After the successful first run of LHC, data taking is scheduled to restart in Summer 2015 with experimental conditions leading increased volumes and event complexity. In order process generated such scenario exploit multicore architectures current CPUs, LHC experiments have developed parallelized software for reconstruction simulation. However, a good fraction their computing effort still expected be executed as single-core tasks. Therefore, jobs diverse resources requirements will...
Leading hybrid and heterogeneous supercomputing systems process hundreds of thousands jobs using complex scheduling algorithms parameters. The centers operating these aim to achieve higher levels resource utilization while being restricted by compliance with policy constraints. There is a critical need for high-fidelity, high-performance tool familiar interfaces that allows not only tuning optimization the operational job scheduler but also enables exploration new management algorithms. We...
Abstract The prompt reconstruction of the data recorded from Large Hadron Collider (LHC) detectors has always been addressed by dedicated resources at CERN Tier-0. Such workloads come in spikes due to nature operation accelerator and special high load occasions experiments have commissioned methods distribute (spill-over) a fraction sites outside CERN. present work demonstrates new way supporting Tier-0 environment provisioning elastically for such spilled-over workflows onto Piz Daint...
This paper introduces the versatile software-defined cluster, a novel framework that integrates HPC and Cloud technologies offering service-oriented approach for computing resources instead of hardware focused one, maintaining infrastructure independence avoiding vendor lock-in. It addresses challenges rigidity lack customizability in conventional systems, facilitating more efficient use shared infrastructures. The core concept revolves around three-tiered structure—Infrastructure, Service...
Complex applications and workflows needs are often exclusively expressed in terms of computational resources on HPC systems. In many cases, other like storage or network not allocatable shared across the entire system. By looking at particular, any workflow application should be able to select both its preferred data manager required capability capacity. To achieve such a goal, new mechanisms introduced. this work, we present tool that dynamically provisions management system top devices. We...
Summary We present a methodology to enable the complete software development life cycle on Cray XC systems within container that can hold any version of Programming Environment (CPE). The installation CPE inside facilitates many aspects typical HPC support and operation workloads managing such as testing new CPEs, comparing performances, or keeping built with an old running updated systems. procedure for creating consists three steps: creation holding targeted CPE, compilation desired...
Complex applications and workflows needs are often exclusively expressed in terms of computational resources on HPC systems. In many cases, other like storage or network not allocatable shared across the entire system. By looking at resource particular, any workflow application should be able to select both its preferred data manager required capability capacity. To achieve such a goal, new mechanisms introduced. this work, we introduce mechanism for dynamically provision management system...
Complex applications and workflows needs are often exclusively expressed in terms of computational resources on HPC systems. In many cases, other like storage or network not allocatable shared across the entire system. By looking at resource particular, any workflow application should be able to select both its preferred data manager required capability capacity. To achieve such a goal, new mechanisms introduced. this work, we introduce mechanism for dynamically provision management system...
As supercomputing systems gradually become an integral part of data driven workflows such as ML and AI or tightly-coupled pre- post-processing pipelines, users need programmable access to shared resources avoid moving large volume dedicated public cloud providers. Public clouds the private ones using technologies like OpenStack, multi-tenancy on hardware has been a commonplace over decade, offering privileged compute, network storage. Such is unavailable batch-scheduled, multi-Petascale...