From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jamal Hadi Salim Subject: Re: [RFC net-next PATCH V2 0/3] qdisc bulk dequeuing and utilizing delayed tailptr updates Date: Thu, 04 Sep 2014 09:44:05 -0400 Message-ID: <54086CA5.6040209@mojatatu.com> References: <20140904125247.4108.8132.stgit@dragon> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8; format=flowed Content-Transfer-Encoding: 7bit Cc: Alexander Duyck , John Fastabend To: Jesper Dangaard Brouer , netdev@vger.kernel.org, "David S. Miller" , Tom Herbert , Eric Dumazet , Hannes Frederic Sowa , Florian Westphal , Daniel Borkmann Return-path: Received: from mail-ie0-f175.google.com ([209.85.223.175]:46187 "EHLO mail-ie0-f175.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750776AbaIDNoN (ORCPT ); Thu, 4 Sep 2014 09:44:13 -0400 Received: by mail-ie0-f175.google.com with SMTP id y20so11718208ier.20 for ; Thu, 04 Sep 2014 06:44:12 -0700 (PDT) In-Reply-To: <20140904125247.4108.8132.stgit@dragon> Sender: netdev-owner@vger.kernel.org List-ID: On 09/04/14 08:54, Jesper Dangaard Brouer wrote: > Wanted people to review this work early... so this is the current > state, even added my debug patch, if people want to "see" it work. > It would be useful to collect some stats when you are experimenting with this stuff on the average batch size the driver sees. Maybe a histogram which shows >1, > 3, > 10 etc that a driver was able to chew on at a time. I realize this may skew your results a little because you have to chew cycles updating such a histogram. I believe that any dumb traffic generator will show lots of packets being batched. But that is not very interesting. It would be hard to show more than 1 in the general case for other types of traffic - but learning what else needs to be tweaked to achive more batch sizes would be helpful. i.e this would help quantify whether the batching is valuable in the general sense... cheers, jamal