This paper studies the problem of routing and scheduling parallel I/O operations to minimize the time required to transfer data between processors and I/O devices. In particular, a 2-dimensional mesh architecture is considered in which routing is performed using wormhole switching and I/O nodes are placed on the periphery of the mesh. Two broad classes of data transfer mechanisms are examined: schedules with blocking (SB), in which packets may be temporarily blocked during transit, and schedules with no blocking (SNB), in which packets are never blocked in the network. For both classes, optimal scheduling is shown to be NP-complete and heuristics are presented and experimentally evaluated via a detailed simulation.
All Science Journal Classification (ASJC) codes
- Theoretical Computer Science
- Hardware and Architecture
- Computer Networks and Communications
- Artificial Intelligence