Using Shared Memory Multiprocessors in High -Performance Networks Protocol Processing

点击翻译按钮获取中文摘要

This research explores and evaluates the use of parallel processing and parallelism in network host protocol processing, through a combination of theoretical analysis and experimental implementation, measurement, and evaluation of a vertically integrated set of host protocol functions. The implementations use Sequent and Silicon Graphics shared memory multiprocessors, connected via FDDI, and a modified MACH operating system kernel. The work considers the protocol functionality of a complete protocol stack, with emphasis on network access, transport, and presentation layer functions.