Choose timezone
Your profile timezone:
Description
As part of the evolution from mainframes to distributed computing CERN developed its own mass storage data management system, SHIFT, which has successfully managed LEP and fixed target data over many years. The requirements of LHC go well beyond the original design of SHIFT and three years ago we adopted a two line approach namely to try a commercial solution, the High Performance Storage System (HPSS) from IBM, and to develop a new Cern Advanced Storage (CASTOR) Manager that would first be used for the next generation of fixed target experiments then extended for LHC.When we started HPSS Compaq Alpha processors were our main tape servers so we included a joint project with a Compaq funded fellow to port the needed parts of HPSS to Compaq while the authors of the SHIFT system took a CERN fellow to work on CASTOR. Both fellows report on their work in this seminar.
1) HPSS 4.1.1 on Compaq Tru64 Unix
by Patrice Calegari, CERN/IT
The High-Performance Storage System (HPSS) provides hierarchical storage management and services for the largest storage environments in the world. It was developed by laboratories and companies in the USA. At CERN, it has been in use since 1998 to store user and experiment data. HPSS evolves quickly and CERN participates in its development on Alpha processors. This presentation will show the improvements that were done this year by porting the last HPSS versions to Compaq Tru64 Unix. An analysis of performance and quality achieved in heterogeneous environments will conclude the presentation.
ABOUT THE SPEAKER
Patrice Calegari, currently working as a fellow in CERN'S IT Division, holds a Magistere (Master's Degree) in Computer Science from the 'Ecole Normale Supérieure de Lyon' and a Ph.D. in Computer Science (parallel computing) from the Swiss Federal Institute of Technology, Lausanne (EPFL). His interests include parallel computing, data management and software engineering.
2) Experience with the CASTOR Mass Storage System: Practical, Design and Development Issues
by Jean-Damien Durand, CERN/IT
This talk will focus on the philosophy behind such a project, the effort and decisions involved in the design (scalability, modularity, HSM functionality, administration, ...), development issues (programming language, portability, appplication-program interface, bindings, ...), and practical issues: how does it work.
ABOUT THE SPEAKER
Jean-Damien Durand, currently working as a fellow in CERN'S IT Division, holds a Ph.D. in Physics. He first got involved in computing with the DELPHI experiment, as software librarian for two years. He then moved to IT-PDP section where he has been working since 1998 on SHIFT and CASTOR.
Organiser(s): L. Pregernig, IT/CE
Organiser(s): L. Pregernig, IT/CE
