CPS and the Fermilab farms
Cooperative Processes Software (CPS) is a parallel programming toolkit developed at the Fermi National Accelerator Laboratory. It is the most recent product in an evolution of systems aimed at finding a cost-effective solution to the enormous computing requirements in experimental high energy physics. Parallel programs written with CPS are large-grained, which means that the parallelism occurs at the subroutine level, rather than at the traditional single line of code level. This fits the requirements of high energy physics applications, such as event reconstruction, or detector simulations, quite well. It also satisfies the requirements of applications in many other fields. One example is in the pharmaceutical industry. In the field of computational chemistry, the process of drug design may be accelerated with this approach. CPS programs run as a collection of processes distributed over many computers. CPS currently supports a mixture of heterogeneous UNIX-based workstations which communicate over networks with TCP/IR CPS is most suited for jobs with relatively low I/O requirements compared to CPU. The CPS toolkit supports message passing remote subroutine calls, process synchronization, bulk data transfers, and a mechanism called process queues, by which one process can find another which has reached a particular state. The CPS software supports both batch processing and computer center operations. The system is currently running in production mode on two farms of processors at Fermilab. One farm consists of approximately 90 IBM RS/6000 model 320 workstations, and the other has 85 Silicon Graphics 4D/35 workstations. This paper first briefly describes the history of parallel processing at Fermilab which lead to the development of CPS. Then the CPS software and the CPS Batch queueing system are described. Finally, the experiences of using CPS in production on the Fermilab processor farms are described.
- Research Organization:
- Fermi National Accelerator Lab., Batavia, IL (United States)
- Sponsoring Organization:
- USDOE, Washington, DC (United States)
- DOE Contract Number:
- AC02-76CH03000
- OSTI ID:
- 10172731
- Report Number(s):
- FNAL/C-92/163; CONF-930117-3; ON: DE92018311
- Resource Relation:
- Conference: 26. Hawaiian international conference on system science: biotechnology computing track,Kauai, HI (United States),5-8 Jan 1993; Other Information: PBD: Jun 1992
- Country of Publication:
- United States
- Language:
- English
Similar Records
Parallel computing for event reconstruction in high-energy physics
Cooperative processes software (CPS)