From: Denys Fedoryschenko Subject: Re: [PATCH] [RESEND] iproute2 : invalid burst/cburst calculation with hrtimers Date: Thu, 5 Feb 2009 11:01:35 +0200 Message-ID: <200902051101.35327.denys@visp.net.lb> References: <200902021926.17768.denys@visp.net.lb> <200902050124.05969.denys@visp.net.lb> <20090205073832.GA4305@ff.dom.local> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Cc: Stephen Hemminger , netdev@vger.kernel.org To: Jarek Poplawski Return-path: Received: from hosting.visp.net.lb ([194.146.153.11]:36594 "EHLO hosting.visp.net.lb" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751707AbZBEJBa (ORCPT ); Thu, 5 Feb 2009 04:01:30 -0500 In-Reply-To: <20090205073832.GA4305@ff.dom.local> Content-Disposition: inline Sender: netdev-owner@vger.kernel.org List-ID: Well, i don't know well what's wrong, but: I have enough powerful router, 2x Quad Core 2 - 2.8 Ghz 4 x e1000e And if i just try to shape interface for 200 Mbps with default iproute2 from git (and latest stable release same situation), i am getting too low burst/cburst. Sometimes it is even miscalculated (it can become 1500 byte, sometimes, while 1600 is minimum) , because yes, Stephen right, it needs float. BUT. Even i set manually twice more burst/cburst, it is just not enough for 200 Mbps. I am checking on Cisco, rate is reaching max 180-185Mbps, and even on this rate buffering in qdiscs and as result jittering hard. So this 100 bytes, which probably missing, really doesn't matter. With 1000 HZ i have class htb 1:10 parent 1:2 rate 220000Kbit ceil 220000Kbit burst 29067b cburst 29067b Sent 14027734219153 bytes 590886842 pkt (dropped 0, overlimits 0 requeues 0) rate 122452Kbit 17131pps backlog 0b 0p requeues 0 lended: 171333348 borrowed: 0 giants: 0 tokens: 980 ctokens: 980 And it works quite well and precise. P.S. Maybe a case, when some guy was not able to reach his gigabit with HTB shaper, i remember there was flowing some discussing, was somehow similar as my case. On Thursday 05 February 2009 09:38:32 Jarek Poplawski wrote: > > I think HZ doesn't matter much in this place, because if it's ever > triggered there is something really wrong and precision is lost. > Anyway, this place was changed a bit in net-next. > > I guess I could misunderstand your previous question. If you mean this > previous thread, then yes, IMHO CONFIG_HZ should matter for precision > (especially places where similar loops use all such jiffies in "normal" > situations). On the other hand, maybe I'm wrong, but using HZ 1000 > with HIGH_RES_TIMERS looks quite obvious to me... > > Jarek P. > -- > To unsubscribe from this list: send the line "unsubscribe netdev" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html