From mboxrd@z Thu Jan 1 00:00:00 1970 From: Daniel Lezcano Subject: Re: [PATCH 4/9] network namespaces: socket hashes Date: Thu, 21 Sep 2006 14:34:53 +0200 Message-ID: <451286ED.4070500@fr.ibm.com> References: <20060815182029.A1685@castle.nmd.msu.ru> <20060815184844.A2882@castle.nmd.msu.ru> <450EB771.9080609@fr.ibm.com> <20060920203222.B10791@castle.nmd.msu.ru> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Cc: netdev@vger.kernel.org, serue@us.ibm.com, haveblue@us.ibm.com, clg@fr.ibm.com, herbert@13thfloor.at, sam@vilain.net, ebiederm@xmission.com, Andrew Morton , dev@sw.ru, devel@openvz.org, alexey@sw.ru Return-path: Received: from mtagate6.uk.ibm.com ([195.212.29.139]:45372 "EHLO mtagate6.uk.ibm.com") by vger.kernel.org with ESMTP id S1751099AbWIUMfF (ORCPT ); Thu, 21 Sep 2006 08:35:05 -0400 Received: from d06nrmr1407.portsmouth.uk.ibm.com (d06nrmr1407.portsmouth.uk.ibm.com [9.149.38.185]) by mtagate6.uk.ibm.com (8.13.8/8.13.8) with ESMTP id k8LCZ3gQ036050 for ; Thu, 21 Sep 2006 12:35:03 GMT Received: from d06av02.portsmouth.uk.ibm.com (d06av02.portsmouth.uk.ibm.com [9.149.37.228]) by d06nrmr1407.portsmouth.uk.ibm.com (8.13.6/8.13.6/NCO v8.1.1) with ESMTP id k8LCbH5v1904650 for ; Thu, 21 Sep 2006 13:37:18 +0100 Received: from d06av02.portsmouth.uk.ibm.com (loopback [127.0.0.1]) by d06av02.portsmouth.uk.ibm.com (8.12.11.20060308/8.13.3) with ESMTP id k8LCZ1A5018707 for ; Thu, 21 Sep 2006 13:35:02 +0100 To: Andrey Savochkin In-Reply-To: <20060920203222.B10791@castle.nmd.msu.ru> Sender: netdev-owner@vger.kernel.org List-Id: netdev.vger.kernel.org Andrey Savochkin wrote: > The main reason is that socket hash tables should be large enough to work > efficiently, but it isn't good to waste a lot of memory for each namespace. > Namespaces should be cheap enough, to allow to have hundreds of them. > This reason of memory efficiency, of course, has a priority unless/until > socket hash tables start to resize automatically. > > Another point is that routing lookup is much more complicated than the > socket's one to add another search key. > Routing also have additional routines for deleting entries matching some > patterns, and so on. > In short, routing is much more complicated, and it already quite efficient > for various sizes of routing tables. That makes sense, thx for the explanation. Cheers. -- Daniel.