Storage resources management at the INFN-CNAF Tier-1

At present we have 2 main mass storage systems for archiving HEP experimental data at the INFN-CNAF Tier-1: a HSM software system (CASTOR) and about 250 TB of different storage devices over SAN. This paper briefly describes our hardware and software environment, and summarizes the technical solutions adopted in order to obtain better availability and high data throughput from the front-end disk servers. In fact, our computing resources, consisting of farms of dual processor nodes (currently about 1000 nodes providing 1300 KspecInt2000), need to access the data through a fast and reliable I/O infrastructure. A valid solution for achieving large I/O throughputs is nowadays provided by parallel file systems. In the last part of this paper some results of detailed tests we performed with GPFS and Lustre over SAN are reported.