{open_quotes}Shift-Betel{close_quotes}: A (very) distributed mainframe

Over the last four years, CERN has progressively converted its central batch production facilities from classic mainframe platforms (Cray XMP, IBM, ESA, Vax 9000) to distributed RISC based facilities, which have now attained a very large size. Both a CPU-intensive system ({open_quotes}CSF{close_quotes}, the Central Simulation Facility) and an I/O-intensive system ({open_quotes}SHIFT{close_quotes}, the Scaleable Heterogeneous Integrated Facility) have been developed, plus a distributed data management subsystem allowing seamless access to CERN`S central tape store and to large amounts of economical disk space. The full system is known as {open_quotes}CORE{close_quotes}, the Centrally Operated Risc Environment; at the time of writing CORE comprises around 2000 CERN Units of Computing (about 8000 MIPs) and over a TeraByte of online disk space. This distributed system is connected using standard networking technologies (IP protocols over Ethernet, FDDI and UltraNet), but which until quite recently were only implemented at sufficiently high speed in the Local Area.