From owner-freebsd-hackers Thu Dec 11 12:14:42 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id MAA11018 for hackers-outgoing; Thu, 11 Dec 1997 12:14:42 -0800 (PST) (envelope-from owner-freebsd-hackers) Received: from gaia.coppe.ufrj.br (cisigw.coppe.ufrj.br [146.164.5.200]) by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id MAA11003 for ; Thu, 11 Dec 1997 12:14:23 -0800 (PST) (envelope-from jonny@coppe.ufrj.br) Received: (from jonny@localhost) by gaia.coppe.ufrj.br (8.8.8/8.8.8) id SAA00694; Thu, 11 Dec 1997 18:13:19 -0200 (EDT) (envelope-from jonny) From: Joao Carlos Mendes Luis Message-Id: <199712112013.SAA00694@gaia.coppe.ufrj.br> Subject: Re: Process scheduling: nice does not work ??? In-Reply-To: <199712100259.TAA02985@usr06.primenet.com> from Terry Lambert at "Dec 10, 97 02:59:11 am" To: tlambert@primenet.com (Terry Lambert) Date: Thu, 11 Dec 1997 18:13:19 -0200 (EDT) Cc: chuckr@glue.umd.edu, tlambert@primenet.com, jonny@coppe.ufrj.br, freebsd-hackers@FreeBSD.ORG X-Mailer: ELM [version 2.4ME+ PL32 (25)] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: owner-freebsd-hackers@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk #define quoting(Terry Lambert) // > > Note: these are base priorities, which the system will adjust based on // ********************************************************************** // > > I/O vs. CPU utilization. // *********************** // // Look at the PRI values in the original posting. The NI values are // irrelevant, no mattery his intent. The NI values should have been used to determine the new PRI values, as have always been in Unix. Maybe the effect of NI differences in FreeBSD is smaller in FreeBSD than in Solaris and Linux ? Is this the right thing ? // If he wanted to "lock" priorities, he need to use rtprio. Otherwise, // the scheduler will drift them as it sees fit. I don't want to lock priorities. I want both processes to run, one with more time ticks than the other. Also, if I put a cpu-intensive process in rtprio, the whole system will starve on CPU. I'll never forget the mess I have done once in Slowlaris when I put the bytebench to run with realtime priority: Nothing else worked: keyboard, ping, etc. Fully frozen. :) Also, never start a rtprio'd file copy over paralell zip drives, as a friend of mine did. :) // IMO, Linux is implementing a "fairness" algorithm based on the NI value; Isn't it what NI for ? If not, what is it for ? // this is not traditional UNIX behaviour. It may favor interactive over // batch response. Note that he was running, effectively, batch processes; // this was my understanding the last time I looked at the Linux scheduler. // // I think Linux is wrong, FWIW. And Solaris also, by peeking at my measures on the first post. Jonny -- Joao Carlos Mendes Luis jonny@gta.ufrj.br +55 21 290-4698 jonny@coppe.ufrj.br Universidade Federal do Rio de Janeiro UFRJ/COPPE/CISI PGP fingerprint: 29 C0 50 B9 B6 3E 58 F2 83 5F E3 26 BF 0F EA 67