Handbook of computer communications standards, Vol. 2: local network standards
Handbook of computer communications standards, Vol. 2: local network standards
PVM: a framework for parallel distributed computing
Concurrency: Practice and Experience
SPLASH: Stanford parallel applications for shared-memory
ACM SIGARCH Computer Architecture News
SIGCOMM '94 Proceedings of the conference on Communications architectures, protocols and applications
TCP/IP illustrated (vol. 2): the implementation
TCP/IP illustrated (vol. 2): the implementation
System support for efficient network communication
System support for efficient network communication
U-Net: a user-level network interface for parallel and distributed computing
SOSP '95 Proceedings of the fifteenth ACM symposium on Operating systems principles
Coordinating parallel processes on networks of workstations
Journal of Parallel and Distributed Computing
Communication Issues in Parallel Computing Across ATM Networks
IEEE Parallel & Distributed Technology: Systems & Technology
High Performance Messaging on Workstations: Illinois Fast Messages (FM) for Myrinet
Supercomputing '95 Proceedings of the 1995 ACM/IEEE conference on Supercomputing
TCP/IP Illustrated: The Protocols v. 1
TCP/IP Illustrated: The Protocols v. 1
Distributed network computing over local ATM networks
IEEE Journal on Selected Areas in Communications
Key Messaging on SOME-Bus clusters
Parallel Computing
A Hybrid Analysis of an Optimization Approach for Cluster Applications
The Journal of Supercomputing
Design and Evaluation of Dynamic Key Message Algorithms for Cluster Computing
HPCASIA '05 Proceedings of the Eighth International Conference on High-Performance Computing in Asia-Pacific Region
| Hi-index | 0.24 |
Networks of workstations (NOWS) are cost-effective platforms to perform parallel computation. Usually, a NOW is not dedicated to parallel jobs. Local users may run some applications in their workstations which involve communications as well. This paper examines the effects of communication interactions of parallel and local jobs on a nondedicated NOW. Three representative communication patterns of parallel jobs are considered. A quantitative model to characterize the interactions is proposed. Measurement results on a NOW support the analytical model and indicate that the network interface in the TCP/IP protocol forms a communication bottleneck during interactions because a standard network interface with a single input/output queue is not able to distinguish communication requests from parallel and local jobs. Therefore, small but important communication messages of a parallel job, such as a barrier synchronization, could be easily blocked by a communication request of a local job, which would degrade the performance of the parallel job significantly. A double queue scheme in the network interface is proposed. Using available information from the protocol layer, the scheme is able to distinguish the two types of communication requests and give a higher priority to parallel jobs' communication requests. The simulation results show that the scheme could improve the performance of parallel jobs without significantly affecting the performance of local jobs.