Warp: an integrated solution of high-speed parallel computing
Proceedings of the 1988 ACM/IEEE conference on Supercomputing
Compiling Fortran D for MIMD distributed-memory machines
Communications of the ACM
Design and Evaluation of primitives for Parallel I/O
Proceedings of the 1993 ACM/IEEE conference on Supercomputing
Architecture and evaluation of a high-speed networking subsystem for distributed-memory systems
ISCA '94 Proceedings of the 21st annual international symposium on Computer architecture
Supporting systolic and memory communication in iWarp
ISCA '90 Proceedings of the 17th annual international symposium on Computer Architecture
A Host Interface Architecture for High-Speed Networks
Proceedings of the IFIP TC6/WG6.4 Fourth International Conference on High Performance Networking IV
Gigabit I/O for distributed-memory machines: architecture and applications
Supercomputing '95 Proceedings of the 1995 ACM/IEEE conference on Supercomputing
A high-speed network interface for distributed-memory systems: architecture and applications
ACM Transactions on Computer Systems (TOCS)
Hi-index | 0.00 |
We consider the problem of high-speed I/O for a single application running on multiple nodes of a distributed-memory parallel computer. Our model is that the parallel system is connected to an I/O system that provides the interface between the internal connections of the parallel system and one or more external connections, such as HIPPI links. We identify two primary operations for this I/O system: scattering data from a high speed link across several lower speed links and gathering data from multiple links onto a single high speed link. We show that these core operations are the basis of the I/O system, independent of the relative speeds of the internal and external connections.We identify several architectural features that are critical for supporting high-speed scatter and gather operations. They include flexible routing methods in the parallel system, low overhead communication, and the ability to support multiple data streams in and out of the memory on the I/O node.