PVM-SCI: A PVM Implementation for Heterogeneous SCI-Clusters
نویسندگان
چکیده
| We present PVM-SCI, an implementation of the PVM 3.4 message passing standard for workstation clusters with scalable coherent interface (SCI). PVM-SCI provides two complementary communication schemes: the usual UDP/IP protocol for messages sent via PVM daemons, and a fast communication protocol that has been implemented directly on top of SCI for messages sent with the PvmRouteDirect directive. No time-consuming protocol stack is used, thereby allowing for a low communication latency and a high message throughput. PVM-SCI is fault-tolerant and it also supports multiple interconnects, giving a convenient and e cient means for utilizing heterogeneous compute clusters. When SCI is available, PVM-SCI makes use of it, otherwise it automatically defaults to the slower Ethernet communication. PVM-SCI runs on Solaris, Linux and Windows NT.
منابع مشابه
Intermediate report on the adaptation of NT - PVM to SCI mechanisms
This report describes the implementation of the Common Message Passing Layer (CML) for the SISCI PC cluster, as far as developed further from the previous report D.2.1.1, and presents the results achieved with this implementation. As CML functionality is integrated into an existing implementation of PVM, we will achieve immediate full functionality of the PVM system over SCI as soon as the amal...
متن کاملParallel Distributed Computing on Sci Workstation Clusters: Early Experiences
Workstation and PC clusters interconnected by SCI (Scalable Coherent Interface) are very promising technologies for high performance cluster computing. Using commercial SBus-SCI interface cards and early system software and drivers, a two-SPARCstation cluster has been constructed for initial testing and evaluation. PVM direct task-to-task data transfers have been adapted to operate on this clus...
متن کاملA Common Messaging Layer for MPI and PVM over SCI
This paper describes the design of a common message passing layer for implementing both MPI and PVM over the SCI interconnect in a workstation or PC cluster. The design is focused at obtaining low latency. The message layer encapsulates all necessary knowledge of the underlying interconnect and operating system. Yet, we claim that it can be used to implement such diierent message passing librar...
متن کاملEmbedding SCI into PVM
The extremely low latencies and high bandwidth results achievable with the Scalable Coherent Interface (SCI) at lowest level encourages its integration into existing Message Passing Environments (MPE's). In combination with Network of Workstations, it can be seen as an alternative for traditional parallel computing with tightly coupled processors with comparable performance at low cost. This pa...
متن کاملCollective Communication on Dedicated Clusters of Workstations
Fast scalable collective operations are very important for achieving good application performance for highly parallel systems. This paper discusses scalability and absolute performance of various algorithms for collective communication. Performance is measured on the world largest SCI cluster, a machine with 96 dual Pentium II nodes interconnected with SCI, showing good scalability and performa...
متن کامل