Advanced Operating Systems   COMP9242 2002/S2

UNSW

Printer-Friendly Version Administration                - Notices - Course Intro - Consultations # On-line Survey (closed) - Survey Results   Work - Lectures - Milestone 0 - Project Admin - Project Spec - Project FAQ - Exam   Documentation - ASysT Lab - L4 source browser - Sulima ISA Simulator - R4x00 ISA Summary  - MIPS R4700 Reference - MIPS R4000 User Manual  - Network Driver - GT64111   Related Info - Aurema OS Prize - OS Hall of Fame   History - 2000 - 1999 - 1998   Staff - Gernot Heiser (LiC)

A Critique of the Critique Data presented earlier: are specific to one (or a few) system, results cannot be generalised without thorough analysis, no such analysis has been done. ==> Cannot trust the conclusions[Lie95]. Re-analysis of Chen & Bershad's Data MCPI for Ultrix and Mach Re-analysis of Chen & Bershad's Data... MCPI caused by cache misses: conflict (black) vs capacity (white) Conclusion  $.$ Mach system (kernel + UNIX server + emulation library) is too big!  $.$ UNIX server is essentially same.  $.$ Emulation library is irrelevant (according to Chan & Bershad). ==> Mach -kernel working set is too big Can we build -kernels which avoid these problems? Requirements for -kernels:  $.$ Fast (system call costs, IPC costs)  $.$ Small (big ==> slow) ==> Must be well designed, providing a minimal set of operations. Can this be done? Are high system costs essential? Example: kernel call cost on i486  ${}$ Mach kernel call: 900 cycles  ${}$ Inherent (hardware-dictated cost): 107 cycles. ==> 800 cycles kernel overhead.  ${}$ L4 kernel call: 123-180 cycles (15-73 cycles overhead). ==> Mach's performance is a result of design and implementation not the -kernel concept!