Abstract :
Since the HERA experiments H1 and ZEUS started data taking in ʹ92, the computing environment at DESY has changed dramatically. Running a mainframe centred computing for more than 20 years, DESY switched to a heterogeneous, fully distributed computing environment within only about two years in almost every corner where computing has its applications.
The computing strategy was highly influenced by the needs of the user community. The collaborations are usually limited by current technology and their ever increasing demands is the driving force for central computing to always move close to the technology edge.
While DESYʹs central computing has a multidecade experience in running Central Data Recording/Central Data Processing for HEP experiments, the most challenging task today is to provide for clear and homogeneous concepts in the desktop area. Given that lowest level commodity hardware draws more and more attention, combined with the financial constraints we are facing already today, we quickly need concepts for integrated support of a versatile device which has the potential to move into basically any computing area in HEP.
Though commercial solutions, especially addressing the PC management/support issues, are expected to come to market in the next 2–3 years, we need to provide for suitable solutions now. Buying PCʹs at DESY currently at a rate of about 30/month will otherwise absorb any available manpower in central computing and still will leave hundreds of unhappy people alone. Though certainly not the only region, the desktop issue is one of the most important one where we need HEP-wide collaboration to a large extent, and right now.
Taking into account that there is traditionally no room for R&D at DESY, collaboration, meaning sharing experience and development resources within the HEP community, is a predominant factor for us.
Keywords :
Data storage & data access , Processor farms , Physics data processing , Mass storage , Network