From owner-freebsd-questions Wed Jun 28 07:44:11 1995 Return-Path: questions-owner Received: (from majordom@localhost) by freefall.cdrom.com (8.6.10/8.6.6) id HAA07900 for questions-outgoing; Wed, 28 Jun 1995 07:44:11 -0700 Received: from plains.nodak.edu (tinguely@plains.NoDak.edu [134.129.111.64]) by freefall.cdrom.com (8.6.10/8.6.6) with ESMTP id HAA07894 for ; Wed, 28 Jun 1995 07:44:09 -0700 Received: (from tinguely@localhost) by plains.nodak.edu (8.6.11/8.6.10) id JAA00651; Wed, 28 Jun 1995 09:43:58 -0500 Date: Wed, 28 Jun 1995 09:43:58 -0500 From: Mark Tinguely Message-Id: <199506281443.JAA00651@plains.nodak.edu> To: freebsd-questions@freebsd.org, rsnow@legend.txdirect.net Subject: Re: Swapinfo again (never decreases) Content-Length: 1172 Sender: questions-owner@freebsd.org Precedence: bulk you say : Step1: swapinfo says 0% used Step2: start X (or any big program) Step3: exit X (or the above mentioned large progra) Step4: swapinfo says 19% used at step2, your RAM is all being used, so some pages are forced to backstore. these pages will stay on backstore until the application for these pages is terminated. A while back I proposed and implemented removing the backstore pages when they were paged back back into RAM. those were the days when *BSD was hanging all the time due to swap depletion. I figure (still do) that things that were paged to backstore may not be the same pages forced back when RAM is depleted again, and we would be prematurely filling the swap and back then my counters determined that it was very seldon occurance that a read backstored page was ever put back into backstore without modifying the page. This idea of removing the backstore page when read from swap never sailed with the VM crew, especially when the VM storing and fetching was improved for FreeBSD. They wanted to keep these pages in swap as a cache. I guess I never fully understood that (not to say they are wrong, I just never heard the reasoning). --mark.