Page 1 of 1

Gargoyle QoS Revised

Posted: Tue Feb 07, 2012 8:54 am
by pbix
Today I upgraded Gargoyle QoS in the repository. The new features will be available in Gargoyle releases after v1.5.3. To get a preview before then you can install any of ispyisail's development builds created after today. I would appreciate your comments if you do. Following are the major new features in this release:

1) Controlled bandwidth sharing by IP within a class. When traffic from/to multiple local IP addresses routes into the same class, QoS will now shares bandwidth equally between them. Previously sharing was not enforced and the results un-even at best. This change will allow large numbers of hosts on your LAN to share bandwidth equally without having to add any rules or classes to your QoS setup.

2) New “Minimize Round Trip Times” attribute in download classes. This provides finer control of exactly when you want to minimize your ping times vs. maximize your WAN link performance.

3) Redesigned active congestion controller switches between MINRTT mode and ACTIVE mode based on which classes are active. The new mode allows maximum WAN link efficiency when MINRTT classes are not active. In ACTIVE mode you will suffer no noticeable decrease in your WAN link speed and still achieve QoS fairness. Adds up to 30% more speed to your downlink vs. the previous version. The new MINRTT mode becomes active when low RTTs are needed and results in lower and more stable ping times than the previous design.

4) The dynamic range of the ACC has been increased from five to one to six to one which allows the ACC to effectively control over a wider range of WAN link bandwidth variation.

Several other minor enhancements are also included in this change including improved accuracy of bandwidth displays shown on QoS pages, improved help text on these pages and several minor JavaScript errors resolved.

These changes are the result of nearly two years of feedback from Gargoyle users on their experiences with Gargoyle QoS including the ACC. Gargoyle remains the only router software with active congestion control and now it is the only router software that can achieve proper QoS operation without requiring the dreaded 30% performance tax on your WAN link.

I started working with Gargoyle as a result of my frustration with the broken QoS I found in DD-WRT. Nearly three years later DD-WRT’s QoS is still non functional while Gargoyle QoS has continuously improved to make it the best gaming router available today. Gargoyle QoS can keep peace in the household by preventing users from interfering needlessly with each other’s online experience.

Experience peace, use Gargoyle! 8-)

Re: Gargoyle QoS Revised

Posted: Tue Feb 07, 2012 12:03 pm
by Eric
Thanks for all your hard work on this, the new version looks really great!

One suggestion: It might be helpful to create a QoS configuration guide on the wiki. I realize that I've been very, very guilty of neglecting the documentation in favor of writing code myself... so if this isn't something you're interested in working on I completely understand. However, given that you've been fairly active in helping people on the forum it may be something you want to consider.

Lately ispyisail has been working on updating the wiki pages on the various sections of Gargoyle based on existing forum posts, but right now the QoS section is rather barren and not very helpful. I think either you or I will have to eventually fill that part in. Again -- no big deal if you'd rather work on other things, but I figured I'd suggest it as something that might increase adoption of Gargoyle QoS, and as something that we can point people to when they ask questions on the forum.

Re: Gargoyle QoS Revised

Posted: Tue Feb 07, 2012 2:15 pm
by pbix
It's on my TODO list. Stay tuned.

Re: Gargoyle QoS Revised

Posted: Tue Feb 07, 2012 6:56 pm
by ispyisail
I wanted to write something about QoS on the wiki but to be honest I don't really understand how to operate it.

I don't have congestion problem with my network and every time I use QoS it it just slows down my network so I just turn it off.

I would suggest that for the average user you need a one click on/off button that just works.

Then for advanced users keep all the advanced settings?

Re: Gargoyle QoS Revised

Posted: Wed Feb 08, 2012 12:01 am
by pbix
Well I agree that the subject of QoS can be complex and will prevent some users from applying it. I wish there was a one button solution but someone smarter than me would need to think of how that could be done. Part of the reason I have not written a Wiki myself is because of how long winded it is likely to be and I wonder if others are likely to read or benefit from it. So I get were you are coming from on this point, just not sure what I can do about it.

Now I do take issue with the other things you stated. First of all everyone experiences congestion. It is a fundamental part of the magic that makes the internet work. Every time you wait for anything to download from the internet you are experiencing congestion and chances are good that it was your WAN link that was congested. It does not matter what type of WAN link you have or how fast it is. Trust me, you experience congestion. Control of congestion can be very beneficial to some users and allows some application to work that otherwise would not but if you do not care that you have to wait then you do not need QoS. Now don't get em wrong, QoS cannot make your WAN link faster but it can make sure that some devices on your LAN get priority over other devices and this can be important.

Second, there is only a small performance penalty for using QoS. With the new revision this penalty will be in the 5% range most of the time. But if you see no benefits then even a 5% penalty is unwanted. Users should understand that QoS does not kill your WAN speed. If you need QoS you will gladly pay the 5%.

Re: Gargoyle QoS Revised

Posted: Mon Feb 13, 2012 12:29 pm
by Markuz Nightwind
Glad to hear that QoS will work even better than before now (thanks Paul :D). Waiting for next ispyisail build to test :))

Re: Gargoyle QoS Revised

Posted: Sat Mar 03, 2012 12:35 pm
by hotzenpl0tz
Been a while since my last post, but I wanted to get this out there: gargoyle's qos implementation has been the primary reason I use the firmware. It is the only solution I have ever found that just completely works for my 3 people appartment, everybody is happy and I haven't had any "the internet is sooo slow" complaints since setting things up. So thanks for your hard work on this, it is really appreciated !

Re: Gargoyle QoS Revised

Posted: Tue Apr 10, 2012 4:00 am
by tals
I started working with Gargoyle as a result of my frustration with the broken QoS I found in DD-WRT. Nearly three years later DD-WRT’s QoS is still non functional while Gargoyle QoS has continuously improved to make it the best gaming router available today. Gargoyle QoS can keep peace in the household by preventing users from interfering needlessly with each other’s online experience.
Speaking from my own experience, prior to Gargoyle I went through a number of routers trying to find something that would keep the family happy with steaming and gaming and failing to the point I believed on adsl it was just not possible to have QOS on the downsteam.

6 months with Gargoyle and the family is happy and the network is running incredibly well - despite the link popping between 7mb-2mb. Great stuff pbix and the Gargoyle team.

Re: Gargoyle QoS Revised

Posted: Wed Jan 20, 2021 2:11 pm
by taz-kun
rookie question!
If the MINRTT is activated when the class is active to optimize according to what I understand, when I am playing the MINRTT is not activated it is as if it needs more than 1 mega of internet to work in this case bf4 uses 120 to 300Kbs MINRTT must be activated or no? From what I understand it should be activated to optimize the game.

Re: Gargoyle QoS Revised

Posted: Fri Jan 22, 2021 5:16 pm
by Lantis
From memory, QoS only goes into active mode if you are utilising more than 12% of your link.
For real-time mode to engage, you need to be in active mode and have a real time class passing data.