Return to BSD News archive
Path: sserve!newshost.anu.edu.au!harbinger.cc.monash.edu.au!msuinfo!agate!howland.reston.ans.net!europa.eng.gtefsd.com!MathWorks.Com!yeshua.marcam.com!charnel.ecst.csuchico.edu!olivea!koriel!male.EBay.Sun.COM!engnews1.Eng.Sun.COM!engnews2.Eng.Sun.COM!attaboy!plocher From: plocher@Sun.COM (John Plocher (x86158 MPk12-3581)) Newsgroups: comp.os.386bsd.misc Subject: Re: dual processor motherboards the way forward? Date: 5 Oct 1994 00:59:44 GMT Organization: Sun Microsystems Inc., Mountain View, CA Lines: 24 Message-ID: <36stq0$177@engnews2.Eng.Sun.COM> References: <Cww6x9.1A0@gnome.co.uk> <36hous$fdv@exile.oec.com> NNTP-Posting-Host: attaboy.eng.sun.com X-Newsreader: TIN [version 1.2 PL2] : : Would it : : be better to just place the kernel on one processor and run all other : : processes on the other; or some form of dynamic load balance which : : will have some overhead? : This sort of evil non-canonical SMP has been attempted by other : vendors before.. This kind of Asymmetric SMP forces all programs to queue on your master cpu for resources whenever they try to do any "kernel" things. For a compute-bound set of processes, this may not matter, but for highly interactive things that use IPC (read/write, sockets, user interfaces, disk I/O, etc) you may end up with a big bottleneck in getting to the master cpu. Of course, this may not be noticable on a system that only has 2 cpus... For comparison's sake, SunOS 4.x uses a form of ASMP on Sun's MP hardware, Solaris 2.x is fully SMP. The complexity difference is large, but the performance differences (scalability wrt # of CPUs...) seem to be worth it :-) -- John Plocher | The code goes in and the graphics come out and nobody really Clusters Group| knows how or why, but computers are all like that. -Ann Gordon