Improving IPC by kernel design

摘要: Inter-process communication (ipc) has to be fast and effective, otherwise programmers will not use remote procedure calls (RPC), multithreading multitasking adequately. Thus ipc performance is vital for modern operating systems, especially μ-kernel based ones. Surprisingly, most μ-kernels exhibit poor performance, typically requiring 100 μs a short message transfer on processor, running with 50 MHz clock rate.In contrast, we achieve 5 μs; twentyfold improvement.This paper describes the methods principles used, starting from architectural design going down coding level. There no single trick obtaining this high performance; rather, synergetic approach in implementation all levels needed. The their synergy are illustrated by applying them concrete example, L3 (an industrial-quality system daily at several hundred sites). main ideas guide complete kernel requirements, make heavy of concept virtual address space inside itself.As experiment shows, significant gains possible: compared Mach, they range factor 22 (8-byte messages) 3 (4-Kbyte messages). Although hardware specific details influence both implementation, these techniques applicable whole class conventional general purpose von Neumann processors supporting addresses. Furthermore, effort required reasonably small, example dedicated parts can concentrated medium sized module.