From owner-freebsd-smp Sun Sep 14 13:03:43 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id NAA27287 for smp-outgoing; Sun, 14 Sep 1997 13:03:43 -0700 (PDT) Received: from sendero-ppp.i-connect.net (sendero-ppp.i-Connect.Net [206.190.143.100]) by hub.freebsd.org (8.8.7/8.8.7) with SMTP id NAA27272 for ; Sun, 14 Sep 1997 13:03:38 -0700 (PDT) Received: (qmail 23500 invoked by uid 1000); 14 Sep 1997 20:03:59 -0000 Message-ID: X-Mailer: XFMail 1.2-alpha [p0] on FreeBSD X-Priority: 3 (Normal) Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 8bit MIME-Version: 1.0 In-Reply-To: <199709140955.CAA14249@usr06.primenet.com> Date: Sun, 14 Sep 1997 13:03:59 -0700 (PDT) Organization: Atlas Telecom From: Simon Shapiro To: Terry Lambert Subject: Re: SMP in FreeBSD 3.x.x Cc: missmanp@milo.cfw.com, freebsd-smp@freebsd.org Sender: owner-freebsd-smp@freebsd.org X-Loop: FreeBSD.org Precedence: bulk Hi Terry Lambert; On 14-Sep-97 you wrote: > > When I worked at that awful place, we actually computed how many > > processors > > one could put on the P6 bus. The number was much, much lower than 30. > > Our number was based on trying to have the CPUs access memory and I/O. > > > With respect, that's because the people designing the access paradigms > weren't very clever. It's possible to allocate from a global pool > to a per processor pool. > > If I'm trying to allocate out of a per processor pool, then I > don't have to contend with other processors to do the allocation. > > In effect, this is a virtual NUMA. > > See the Dynix paper: > > Efficient Kernel Memory Allocation on Shared Memory Multiprocessors > McKenney, P.E. and Swignline , J. > Proceeding of Winter 1993 Usenix Technical Conference > Jan 1993, pages 295-303 The Dynix papers were used as a model to scale SMP vs. MPP. The concusion (in another job, the awful place did not know their head from their tail :-) was that aroud 30 processors is where scalability will fall off. We used Dynix + Oracle for O/S application model. Processor was Pentium. If I remember correctly, the P6-200 has worse instructions/memory/IO bandwidth ratios than Pentiums-66 does. That led to the conclusion that we will not grow beyond 30 either. I will not go into what the initial P7 was supposed to do. I have no clue where it ended being. --- Sincerely Yours, (Sent on 14-Sep-97, 12:58:09 by XF-Mail) Simon Shapiro Atlas Telecom Senior Architect 14355 SW Allen Blvd., Suite 130 Beaverton OR 97005 Shimon@i-Connect.Net Voice: 503.643.5559, Emergency: 503.799.2313