From: don-lartc@isis.cs3-inc.com (Don Cohen) Date: Thu, 09 May 2002 15:47:23 +0000 Subject: [LARTC] Re: More on qdiscs - testing packet timeout Message-Id: List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: lartc@vger.kernel.org Implement your idea and measure difference in delay, jitter, thoughtput of TCP and reaction to flooding. In other words show that your idea can do better that current approach. When I was speking about idea like HTB nobody listened. I've to implement it .. It is implemented and I do use it. But under normal circumstances it has no effect. Normally packets don't stay in the queue for very long. I can show you cases where it does help, similar to the example I sent. A more interesting question is under what circumstances would it have a bad effect. But I don't think that question can be answered by just trying a few tests. Possibly long term use, or use by lots of different people (in different situations) would turn up something, but I think its best to start by just trying to think of such cases. In the mean while, if you'd like me to run a few examples that show no difference for normal things, that would be easy enough, and similarly I could show an example where a queue that is, in essence, too long for its low rate has the effect of denying service for a long time with a long packet timeout and for a short time with a short packet timeout. I'm thinking in terms of a timeout of a few seconds. _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/