From owner-freebsd-stable Sun Sep 23 7:47:12 2001 Delivered-To: freebsd-stable@freebsd.org Received: from lurza.secnetix.de (lurza.secnetix.de [212.66.1.130]) by hub.freebsd.org (Postfix) with ESMTP id 95F5737B422 for ; Sun, 23 Sep 2001 07:47:08 -0700 (PDT) Received: (from olli@localhost) by lurza.secnetix.de (8.9.3/8.9.3) id QAA47034; Sun, 23 Sep 2001 16:47:00 +0200 (CEST) (envelope-from oliver.fromme@secnetix.de) Date: Sun, 23 Sep 2001 16:47:00 +0200 (CEST) Message-Id: <200109231447.QAA47034@lurza.secnetix.de> From: Oliver Fromme To: freebsd-stable@FreeBSD.ORG, nuzrin@goose.net.my Reply-To: freebsd-stable@FreeBSD.ORG, nuzrin@goose.net.my Subject: Re: options HZ=1000 In-Reply-To: <200109231401.WAA03886@venus.cyber.mmu.edu.my> X-Newsgroups: list.freebsd-stable User-Agent: tin/1.5.4-20000523 ("1959") (UNIX) (FreeBSD/4.1-RELEASE (i386)) MIME-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8bit Sender: owner-freebsd-stable@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG nuzrin yaapar wrote: > I've been searching for quite some time now but cannot find any pointers. > It's regarding the options HZ=1000 in the kernel configuration file. Do > anyone know what impact this setting might have to the overall performance? > Any other things I should be aware of when I use this options? > > Any experiences shared or pointers will be greatly appreciated. It depends what your intention is. To increase scheduler granularity, I'd recommend not to fiddle with HZ, but instead to decrese the kern.quantum sysctl. This helps especially when you're running processes that are cpu-bound and processes that are i/o- bound on the same machine. I've seen this on fileservers running rc5 (or seti, or prime number crunchers, or similar stuff). As soon as you started the cpu-bound process, the network performance of the fileserver dropped down noticeable, no matter whether you used nice (or even idprio) or not. Tuning kern.quantum fixed the problem. Increasing the rate by a factor of 10 shouldn't have any significant overhead. Keep in mind that a 1 GHz processor executes one million cycles within one millisecond. So, one millisecond is quite an eternity for a CPU. :-) Regards Oliver -- Oliver Fromme, secnetix GmbH & Co KG, Oettingenstr. 2, 80538 München Any opinions expressed in this message may be personal to the author and may not necessarily reflect the opinions of secnetix in any way. "All that we see or seem is just a dream within a dream" (E. A. Poe) To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message