From owner-freebsd-questions Mon Feb 16 02:07:45 1998 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id CAA12712 for freebsd-questions-outgoing; Mon, 16 Feb 1998 02:07:45 -0800 (PST) (envelope-from owner-freebsd-questions@FreeBSD.ORG) Received: from gwdu60.gwdg.de (gwdu60.gwdg.de [134.76.10.60]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id CAA12695 for ; Mon, 16 Feb 1998 02:07:38 -0800 (PST) (envelope-from kheuer@gwdu60.gwdg.de) Received: from localhost (kheuer@localhost) by gwdu60.gwdg.de (8.8.5/8.8.5) with SMTP id LAA04107; Mon, 16 Feb 1998 11:06:35 +0100 (CET) Date: Mon, 16 Feb 1998 11:06:35 +0100 (CET) From: Konrad Heuer To: Greg Lehey cc: John Kelly , John Goerzen , Vincent Defert , freebsd-questions@FreeBSD.ORG Subject: Re: FreeBSD vs Linux In-Reply-To: <19980215160519.03942@freebie.lemis.com> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-questions@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG On Sun, 15 Feb 1998, Greg Lehey wrote: > On Sun, 15 February 1998 at 6:18:06 +0000, John Kelly wrote: > > On Sat, 14 Feb 1998 22:38:12 -0600 (CST), John Goerzen > > wrote: > > > >> On Thu, 12 Feb 1998, Konrad Heuer wrote: > >> > >>> 1. The Linux scheduler which is very different from other UNIX > >>> schedulers (and thus the FreeBSD scheduler) behaves very poor when the > >>> system is heavily loaded (no fair scheduling!). > >> > >> In comparing Linux to SunOS and Solaris in heavliy-loaded systems, I can > >> say that Linux performed much better. However, I have not compared it > >> directly to FreeBSD. > > > > I have. Linux comes apart at the seams under load. FreeBSD just > > keeps going and going and going .... > > This ties in with just about every report I've heard about Linux, > though I have no personal experience in the area. I also find it hard > to believe that Linux should be able to beat Solaris 2 in this area. To go more into details: I didn't want to say the the interactive response times of heavily loaded Linux systems are bad in general. The more serious problem is that the Linux dynamic process priorities are in fact cpu times the process will be able use within a time interval of unknown length. After consuming all of its `priority stock' a process will *not* be started again until there's no more runnable process in the system with a non-zero priority. Then and only then each such process is assigned again its original static priority which is based on the nice value, and the game starts all over again. Therefore, in case of heavily loaded systems, there's a high risk that the scheduling will not be fair. The FreeBSD scheduler as many UNIX schedulers also punishes a cpu-bound process by increasing its dynamic process priority value and thus lowering the priority. But on the other hand it uses a decay filter to reduce the priority value periodically to give such a process a fair chance again after a definite number of clock ticks. K. Heuer, GWDG, Goettingen, Germany (kheuer@gwdu60.gwdg.de) To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message