Interoperating Cloud-based Virtual Farms

21st International Conference on Computing in High Energy and Nuclear Physics

物理学;计算机科学

Bagnasco, S.^1 ; Colamaria, F.^2 ; Colella, D.^2 ; Casula, E.^3 ; Elia, D.^4 ; Franco, A.^4 ; Lusso, S.^1 ; Luparello, G.^5 ; Masera, M.^7 ; Miniello, G.^2 ; Mura, D.^3 ; Piano, S.^6 ; Vallero, S.^7 ; Venaruzzo, M.^8 ; Vino, G.^2

INFN Sezione di Torino, Italy^1

Università degli Studi di Bari, INFN Sezione di Bari, Italy^2

Università degli Studi di Cagliari, INFN Sezione di Cagliari, Italy^3

INFN Sezione di Bari, Italy^4

Università degli Studi di Trieste, INFN Sezione di Trieste, Italy^5

INFN Sezione di Trieste, Italy^6

Università degli Studi di Torino, INFN Sezione di Torino, Italy^7

INFN Laboratori di Legnaro, Italy^8

关键词: Analysis softwares;    Computing resource;    Distributed analysis;    Distributed clouds;    Distributed storage;    End-user perspective;    Interactive analysis;    Storage capacity;   

Others  :  https://iopscience.iop.org/article/10.1088/1742-6596/664/2/022033/pdf
DOI  :  10.1088/1742-6596/664/2/022033

学科分类：计算机科学（综合）

来源: IOP

【 摘 要 】

The present work aims at optimizing the use of computing resources available at the grid Italian Tier-2 sites of the ALICE experiment at CERN LHC by making them accessible to interactive distributed analysis, thanks to modern solutions based on cloud computing. The scalability and elasticity of the computing resources via dynamic ("on-demand") provisioning is essentially limited by the size of the computing site, reaching the theoretical optimum only in the asymptotic case of infinite resources. The main challenge of the project is to overcome this limitation by federating different sites through a distributed cloud facility. Storage capacities of the participating sites are seen as a single federated storage area, preventing the need of mirroring data across them: high data access efficiency is guaranteed by location-aware analysis software and storage interfaces, in a transparent way from an end-user perspective. Moreover, the interactive analysis on the federated cloud reduces the execution time with respect to grid batch jobs. The tests of the investigated solutions for both cloud computing and distributed storage on wide area network will be presented.