Distributed cloud computing in high energy physics

Abstract

Cloud computing is increasingly being used for running high energy physics (HEP) applications. We review the motivation for using clouds in HEP and describe how they are gradually being integrated into our systems. In particular, we highlight our use of a distributed cloud computing system that integrates both private and public IaaS clouds into a unified infrastructure. We describe our experience using the distributed cloud and our plans to make the system context-aware in order to scale to larger workloads and run data-intensive HEP applications.

Context-aware cloud computing for HEP

“…For example, the Belle II experiment at the KEK Laboratory in Japan uses four different methods to exploit clouds [4]. This paper will focus on our system that integrates private and opportunistic clouds into a single batch computing environment [5].…”

“…Our system has been in production operation for many years using clouds in North America, Australia and Europe, completing many millions of jobs for ATLAS and Belle II [5]. The number and types of clouds has evolved due to the rapid development in cloud technologies; for example, the commonly used OpenStack software did not exist when CloudScheduler was developed.…”

“…We have been using distributed cloud centres for HEP workloads [5]. Many of the clouds we use are opportunistic resources, hosted by research institutions that are not involved in HEP research or commercial clouds.…”

Μελέτη Υπολογιστικών Μεθόδων Για Παράλληλη Προσομοίωση Του Βαρυτικού Προβλήματος Ν-Σωμάτων

“…Computations and many studies have been proposed in order to discuss the potential of HPC on Cloud environments, [33,49,159]. Further, many applications and a variety of physics simulations have started to migrate to Cloud environments, [127,143], such as the Ansys and Solidworks software environments. Recent studies suggest new, more efficient and effective, service delivery models for HPC applications, such as the CloudLightning-Resource-as-a-Service (CL-RaaS), [O3], or Workflow-as-a-Service (WaaS), [125], that complement the classic service delivery models.…”

“…Many HPC applications and a variety of physics simulations have started to migrate to Cloud environments, [127,143]. Examples of simulation software include the Ansys and Solidworks environments.…”