Gargoyle Forum

Thanks very much. That was actually an excellent explanation for laymen. I appreciate you taking the time.

+1

When internet speeds were much slower it probably didn't matter so much

Things are changing

Lantis wrote: Now, why is Gargoyle's QoS having such an impact on performance? My theory is that the maximum lengths of the queues is not enough. There's some maths behind it which i haven't understood yet (and i've reached out to someone who knows, or at least used to), but the theoretical maximum through a single queue is essentially 350Mbps.
It might be as simple as lengthening the queue, or we might need to introduce additional queues, or maybe SFQ is not appropriate above a certain bandwidth limit. I'm not sure yet, but i'm looking.

Did this ever get followed up on? That is did anything change with Gargoyle's QoS?

Obviously the Gargoyle developers figured they had a better way to do it or was simply using disciplines like Cake from openwrt harder to implement?

That is did anything change

Yes, the internet got faster and the hardware is having problem keeping up.

ispyisail wrote:

That is did anything change

Yes, the internet got faster and the hardware is having problem keeping up.

I was referring to Landis' post where he mentions why he thinks QoS in Gargoyle causes so much more of a performance hit than Cake does in Openwrt. He offered up a possible fix, lengthening the queue, but I have no idea if it was ever tried.

I don't have nearly the same performance hit when using QoS in either openwrt or dd-wrt which recently added Cake.

This wasn't a complaint. I was just curious if it had gotten better so I might try Gargoyle again.

Cake wasn't around when Gargoyles QOS was implemented (I think)

The two implementations are quite different and let you achieve different things.
Gargoyle is very good at limiting individual flows, whereas cake is better at just sharing a link fairly.

I did look into the queues, and my tests showed no difference.
I need to run the tests again on slower hardware as it may turn out they make a bigger difference the worse your hardware is.

Lantis wrote:Cake wasn't around when Gargoyles QOS was implemented (I think)

The two implementations are quite different and let you achieve different things.
Gargoyle is very good at limiting individual flows, whereas cake is better at just sharing a link fairly.

Would it be possible to add cake or fq_codel as choices for QOS in Gargoyle. Along side the version Gargoyle uses? I was going to ask in the "Features requests" forum but thought I'd follow up here first to see if it was even possible.

I'm using dd-wrt right now but have grown tired of it never having milestone or release builds like open-wrt and Gargoyle do. Or most all software for that matter. I don't mind beta testing but at some point I believe software needs to have it's features set and just work on fixing bugs before moving on to the next version.

I don't think they will ever be selectable from the GUI, but you can swap fq_codel in Gargoyles qos implementation with a few lines of code (there's an example in /etc/unit.d/qos_gargoyle)

I don't think cake will be done.

Lantis wrote:I don't think they will ever be selectable from the GUI, but you can swap fq_codel in Gargoyles qos implementation with a few lines of code (there's an example in /etc/unit.d/qos_gargoyle)

I don't think cake will be done.

ah. Interesting. Thanks. Does this file from pbix still work in the current 1.12.0 version?
viewtopic.php?t=5726

No, i wouldn't have thought so.

You can see the instructions and line from pbix here:
https://github.com/ericpaulbishop/gargo ... #L460-L471
This is in the latest code.

You would swap line 458 with line 471.
As noted, you need to make a similar modification on the Download side, and no instructions are provided for that. "It is left to the student".

As also noted in the comments, at least previously, no noted improvements (and in fact, some stability issues crept in) when switching, but please feel free to experiment as you wish and report back.