From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S261790AbUBWDdZ (ORCPT ); Sun, 22 Feb 2004 22:33:25 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S261794AbUBWDdZ (ORCPT ); Sun, 22 Feb 2004 22:33:25 -0500 Received: from mail.aei.ca ([206.123.6.14]:57042 "EHLO aeimail.aei.ca") by vger.kernel.org with ESMTP id S261790AbUBWDdY (ORCPT ); Sun, 22 Feb 2004 22:33:24 -0500 From: Ed Tomlinson Organization: me To: linux-kernel@vger.kernel.org Subject: Re: Large slab cache in 2.6.1 Date: Sun, 22 Feb 2004 22:33:20 -0500 User-Agent: KMail/1.5.93 Cc: Mike Fedyk References: <4037FCDA.4060501@matchmail.com> <200402220903.08299.edt@aei.ca> <40396551.9030002@matchmail.com> In-Reply-To: <40396551.9030002@matchmail.com> MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <200402222233.20426.edt@aei.ca> Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On February 22, 2004 09:28 pm, Mike Fedyk wrote: > Ed Tomlinson wrote: > > On February 21, 2004 10:28 pm, Linus Torvalds wrote: > >>On Sat, 21 Feb 2004, Chris Wedgwood wrote: > >>>Maybe gradual page-cache pressure could shirnk the slab? > >> > >>What happened to the experiment of having slab pages on the (in)active > >>lists and letting them be free'd that way? Didn't somebody already do > >>that? Ed Tomlinson and Craig Kulesa? > > > > You have a good memory. > > > > We dropped this experiment since there was a lot of latency between the > > time a slab page became freeable and when it was actually freed. The > > current call back scheme was designed to balance slab preasure and > > vmscaning. > > Does it really matter if there is a lot of latency? How does this > affect real-world results? IOW, if it's not at the end of the LRU, then > there's probably something better to free instead... It mattered. People noticed and complained. In any case, as Andrew pointed out, we get the same effect, without long latencies, in a simplier manner with the current scheme. Ed