Production control system: A resource management and load leveling system for mainframe computers

摘要

The Livermore Computing Production Control System, commonly called the PCS is described in this paper. The intended audiences for this document are system administrators and resource managers of computer systems. In 1990, the Livermore Computing Center committed itself to convert its supercomputer operations from the New Livermore TimeSharing System (NLTSS) to UNIX based systems. This was a radical change for Livermore Computing in that over thirty years had elapsed during which the only operating environments used on production platforms were LTSS and then later NLTSS. Elaborate, and sometimes obscure, paradigms (to others) were developed to help the lab's scientists productively use the machines and to accurately account for their use to government oversight agencies. Resource management is difficult in typical UNIX and related systems. The weakness involves the allocation, control and delivery of the usage of computer resources. This is a result of the origins and subsequent development of UNIX which started as a system for small platforms in a collegial or departmental atmosphere without stringent or complex budgetary requirements. Accounting also is weak, being generally limited to reporting that a user used an amount of time on a process. A process accounting record is made only after the completion of a process and then only if the process does not terminate as a result of a system (open quotes)panic.(close quotes)