From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S263561AbTJQR0c (ORCPT ); Fri, 17 Oct 2003 13:26:32 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S263562AbTJQR0c (ORCPT ); Fri, 17 Oct 2003 13:26:32 -0400 Received: from gw1.cosmosbay.com ([62.23.185.226]:31466 "EHLO gw1.cosmosbay.com") by vger.kernel.org with ESMTP id S263561AbTJQR02 (ORCPT ); Fri, 17 Oct 2003 13:26:28 -0400 Message-ID: <02b901c394d3$96c662e0$890010ac@edumazet> From: "dada1" To: "David S. Miller" Cc: , , References: <1066356438.15931.125.camel@cube><20031017023437.GB28158@work.bitmover.com><01e601c39484$f3fa31c0$890010ac@edumazet> <20031017021040.4964309a.davem@redhat.com> Subject: Re: /proc reliability & performance Date: Fri, 17 Oct 2003 19:24:56 +0200 MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 6.00.2800.1158 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1165 Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org From: "David S. Miller" > "dada1" wrote: > > > A "cat /proc/net/tcp" takes too much time to even try it. :( > > > > tools like "netstat" or "lsof", (even with -n flag) are just unusable. > > Because they don't use the netlink TCP socket dumping > facility which is made to handle such things much better > than procfs ever can. Thanks David for the hint. :) I buy it. I found that the ss command from iproute2 package does use the 'netlink TCP dumping' you mention (how many people on earth heard about that ?) Instead of 15 minutes for a 'netstat -n > FILE', my server takes now 6 seconds with 'ss -n > FILE', with 200000 sockets opened. Eric