[Cerowrt-devel] CeroWrt 3.10 AQM page

Sebastian Moeller moeller0 at gmx.de
Sat Dec 21 05:01:41 EST 2013


Rich Brown <richb.hanover at gmail.com> wrote:
>Folks,
>
>I have updated the CeroWrt 3.10 AQM page. Thanks for all the comments,
>I’ll incorporate more comments as people send them in. Some thoughts on
>the page so far:
>
>- I agree that we should keep the descriptions generic (that is, not
>tailored specifically to CeroWrt) so we can push into OpenWrt without
>changes. 
>
>- It’s also a good idea to look for good marketing name (SQM, IQM, etc)
>so that we can tout the goodness of the fq_codel, &c. I’ll submit
>another proposal on the “Anything but ‘AQM’” thread.

Could any of us come up with a decent backronym?


>
>- I removed references to calling the ISP - sometimes it could work,
>but I see now how it could also be problematic.

      Too bad since the ISP will have this information available, the alternatives are Google and luck or long measurements...

>
>- I prefer telling people to decide on their link layer adaptation
>using the term “ADSL” instead of “ATM” (even though it’s technically
>more true) since nobody knows what ATM is, and they (most likely) do
>know whether they have DSL, Cable, Fiber, or something else

      I think ATM is the way to go, otherwise those unlucky souls on VDSL links with ATM carriers will be quite confused. And in case ADSL should learn to use PTM instead the GUI and wiki will still be okay... Assuming the users are willing to learn ATM might work...


>
>- I wasn’t clear where a decision about overhead of 40 or 44 comes
>from, or where it should be described

     44 seems to be worst case overhead for DSL over atm , so defaulting to 44 would work for everybody, wasting some bytes per packet for most. 40 seems to be the largest size that is actually common, wasting less for everyone... Both of these are with PPPoE. I had a look at tomato's and gargoyle, one of the tomato's let's the user select the encapsulation from a long list, the other one automatically selects link layer ADSL when pppoe is in use. I argue that both methods are suboptimal, the former assumes quite a bit of information from the user, the latter has too many false positives (in Germany vdsl2 typically uses pppoe but PTM).


>
>- Are there any better (but still easy to use) speed test tool besides
>speedtest.net?

      I think these tools pretty much all are not too helpful for our purpose as they measure average bandwidth over the whole path, while we need slowest non-shared link, aka the bottleneck we can actually reliably affect with sqm

>
>- It’s a good idea to make it easy for people to get back to defaults.
>Could there be a “Restore Defaults” button on the Basic Settings tab? 


      Not sure, the queueing setup script selection is quite nonstandard...
>
>Draft #2 of the Setting up AQM page is available at:
>http://www.bufferbloat.net/projects/cerowrt/wiki/Setting_up_AQM_for_CeroWrt_310
>
>Thanks in advance for all your comments.

      Thanks for creating this, the next version of the GUI will contain a link. It would be excellent if we could have a new name by then...

>
>Rich
>
>PS And if you’re running short of things to work on, the new “Quick
>Test for Bufferbloat” page *really* needs help. 

Hi Rich,
-- 
Sent from my Android phone with K-9 Mail. Please excuse my brevity.



More information about the Cerowrt-devel mailing list