Abstract

The mass storage challenge for the experiments at the Large Hadron Collider (LHC) is still nowadays a critical issue for the various Tier-1 computing centres and the Tier-0 centre involved in the custodial and analysis of the data produced by the experiments. In particular, the requirements for the tape mass storage systems are quite strong, amounting to about 15 PB of data produced annually that should be available for near-line access at any time. Besides the solutions already widely employed by the High Energy Physics community so far, INFN-CNAF has approched, in the last year, a solution based on the collaboration between the General Parallel File System (GPFS) and the Tivoli Storage Manager (TSM) by IBM and StoRM as Storage Resource Management (SRM) interface, developed at INFN. The new features available in GPFS version 3.2 allow in general to interface GPFS with any tape storage manager. We implemented such an interface for TSM, and we performed various performance studies on a prototype testbed system. The first StoRM/GPFS/TSM based system is already in production at CNAF for T1D1 Storage Class; it is used by the LHCb experiment. Currently we are performing new tests to exploit features implemented in the new versione of TSM 6.1. These features are focused in high performance and optimized access to the tape backend. We will describe the implementation of the interface and details of the prototype testbed for T1D0 SC and we will discuss the results of the LHCb production system.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call