Skip to Main Content
The new era of particle physics poses strong constraints on computing and storage availability for data analysis and data distribution. The SuperB project plans to produce and analyzes bulk of dataset two times bigger than the actual HEP experiment. In this scenario one of the main issues is to create a new cluster setup, able to scale for the next ten years and to take advantage from the new fabric technologies, included multicore and graphic programming units (GPUs). In this paper we propose a new site-wide cluster setup for Tier1 computer facilities, aimed to integrate storage and computing resources through a mix of high density storage solutions, cluster file system and Nx10Gbit/s network interfaces. The main idea is overcome the bottleneck due to the storage-computing decoupling through a scalable model composed by nodes with many cores and several disks in JBOD configuration. Preliminary tests made on 10Gbit/s cluster with a real SuperB use case, show the validity of our approach.
Date of Conference: 21-24 June 2011