From owner-freebsd-hackers Wed Sep 24 07:57:44 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id HAA03058 for hackers-outgoing; Wed, 24 Sep 1997 07:57:44 -0700 (PDT) Received: from labinfo.iet.unipi.it (labinfo.iet.unipi.it [131.114.9.5]) by hub.freebsd.org (8.8.7/8.8.7) with SMTP id HAA03048 for ; Wed, 24 Sep 1997 07:57:38 -0700 (PDT) Received: from localhost (luigi@localhost) by labinfo.iet.unipi.it (8.6.5/8.6.5) id PAA21711 for hackers@freebsd.org; Wed, 24 Sep 1997 15:41:37 +0200 From: Luigi Rizzo Message-Id: <199709241341.PAA21711@labinfo.iet.unipi.it> Subject: disk scheduling... To: hackers@freebsd.org Date: Wed, 24 Sep 1997 15:41:37 +0200 (MET DST) X-Mailer: ELM [version 2.4 PL23] Content-Type: text Sender: owner-freebsd-hackers@freebsd.org X-Loop: FreeBSD.org Precedence: bulk I seem to remember this was debated some time ago but do not remember the status of the code. At any rate.. on a 2.2.1 system (SCSI for what matters), I am doing moving a large subtree from one position to another, and it turns out that during this operation interactive performance is awful for data which is not cached. I think to remember that the explaination was that there is no fair sharing of disk access, so a process queueing large blocks of io can eat almost all the resources. I am not sure if the problem can be solved easily (in that at some point transactions might have become anonymous...) Still if I wanted to look at possible solutions, can someone point me to the right place (documentation, sources, hints etc. ?) Perhaps some VM guru can help ? Thanks Luigi