Menu
▾
▴
Re: [Astlinux-devel] traffic-shaper: fq_codel (Fair Queueing CoDel)
|
From: Michael K. <mic...@ip...> - 2017-06-01 23:50:13
|
Thanks Lonnie. Im looking forward to testing it out. Regards Michael Knill -----Original Message----- From: Lonnie Abelbeck <li...@lo...> Reply-To: AstLinux Developers Mailing List <ast...@li...> Date: Friday, 2 June 2017 at 9:46 am To: AstLinux Developers Mailing List <ast...@li...> Subject: Re: [Astlinux-devel] traffic-shaper: fq_codel (Fair Queueing CoDel) Hi Michael, Basically yes, though this shaping stuff seems more of an art than a science. The 'htb' has some built-in reservation for priority traffic, for my 10 Mbps up it had a little too much reserved so I ended up setting the Uplink speed to 12000, though if you had a lower speed and a lot of voice traffic you might set it to less than the measured uplink speed. For the 'hfsc' shaper it is somewhat more dynamic, so setting it to the uplink speed is a good first guess (my ISP uplink is measured a little over 10 Mbps), but if you had a lot of voice traffic you may need to bump it down somewhat. For example using the 'hfsc' shaper, I maxed out the uplink with a HTTP speed test and ran a SIP call out and back, you could see the up max-speed test drop (minus one side of the SIP traffic) while the SIP call was smooth. Regardless it takes a little trial and error testing. Lonnie On Jun 1, 2017, at 5:49 PM, Michael Knill <mic...@ip...> wrote: > Hi Lonnie > > Thanks again for all your work. > > So with htb you shape default traffic to the Line Rate (or line CIR) minus Required Voice Bandwidth (e.g. 100K G.711 x max sessions). But with hfsc, you set it to the Line Rate (or line CIR) and default traffic has access to the full Line Rate while there is no voice traffic. Is this correct? > > Regards > Michael Knill > > From: Lonnie Abelbeck <li...@lo...> > Reply-To: AstLinux Developers Mailing List <ast...@li...> > Date: Friday, 2 June 2017 at 8:31 am > To: AstLinux Developers Mailing List <ast...@li...> > Subject: [Astlinux-devel] traffic-shaper: fq_codel (Fair Queueing CoDel) > > Hi Devs, > > A quick note, as of r8362, the "traffic-shaper" plugin, now uses the fq_codel (Fair Queueing CoDel) for both 'htb' and 'hfsc' types. > > In the past, many of us have found the 'hfsc' shaper not as good as expected. With the addition of fq_codel the 'hfsc' shaper type works much better than before in my limited testing. > > Personally I have a 50/10 Mbps cable connection, so I set the Uplink Speed to 10000 and disabled any ingress shaping, ex. > > PastedGraphic-1.tiff > > Testing this sort of shaping stuff is difficult to do, but so far is working well for me. > > Everyone reading this should do some tests of their own. > > Lonnie > > ------------------------------------------------------------------------------ > Check out the vibrant tech community on one of the world's most > engaging tech sites, Slashdot.org! http://sdm.link/slashdot_______________________________________________ > Astlinux-devel mailing list > Ast...@li... > https://lists.sourceforge.net/lists/listinfo/astlinux-devel ------------------------------------------------------------------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, Slashdot.org! http://sdm.link/slashdot _______________________________________________ Astlinux-devel mailing list Ast...@li... https://lists.sourceforge.net/lists/listinfo/astlinux-devel |
