ACC issue

Report issues relating to bandwith monitoring, bandwidth quotas or QoS in this forum.

Moderator: Moderators

Post Reply
bawjkt
Posts: 32
Joined: Tue Jul 03, 2012 1:11 pm

ACC issue

Post by bawjkt »

Using 1.5.5 at a bar. 12mbit link.

Today link was limited to ~2 Mbit.

Without router, link returned 11.5 Mbit.

So I checked DL QOS page.

Initially I had been running QOS with ACC on upload and download at the measured rates. Like I do in every other environment; it's the best.

Today, for whatever reason, Gargoyle in this deployment measured pingtime to ISP (stock) gateway at 400+ms, which resulted in ACC throttling back the 12Mbit connection to 2Mbit, trying to fight the latency. I was notified, saw it, and fixed it.

I changed the ping target to a local ISP DNS server, typically 15ms away.
This fixed the problem for a few minutes, but it soon shot up to over the pingtime limit and ACC kicked in again, throttling to "fair limits" of a small percentage of link speed.

So ultimately I shut all QOS off and the line returned to 11.5Mbit speeds. Did I have time (or forethought) to measure actual latency to ISP gateway or alternate target before I shut QOS off? No.

And really, no matter what the latency, if it comes back to 97% of advertised rate (from 16% throttled by ACC seeing 400+ms) within seconds after shutting down all QOS, you'll do that.

Anyway, wherever this artifact came from, it resulted in the linespeed being throttled by 85%.

So I'm reporting my first ever problem with Gargoyle.

I can easily recreate the situation; I just don't have alarms in place to tell me when the connection gets throttled to 2Mbit again. This happened for a few days. How/why could the router observe 400ms pingtimes to the ISP gateway via cable modem not under load?

What kind of ISP gateway is over 400ms away? Cable modem? Get real.

We don't know. The consequent throttling we do know about.

Pbix login details in private message...

pbix
Developer
Posts: 1371
Joined: Fri Aug 21, 2009 5:09 pm

Re: ACC issue

Post by pbix »

bawjkt wrote: How/why could the router observe 400ms pingtimes to the ISP gateway via cable modem not under load?


This is the heart of the issue and to really understand what happened you will need to find out the answer to this question.

For some applications a 400ms ping would lead to poor performance. However, if you want to use a longer ping limit due to the nature of your ISP connection you can enter one in the ACC configuration section.
Linksys WRT1900ACv2
Netgear WNDR3700v2
TP Link 1043ND v3
TP-Link TL-WDR3600 v1
Buffalo WZR-HP-G300NH2
WRT54G-TM

bawjkt
Posts: 32
Joined: Tue Jul 03, 2012 1:11 pm

Re: ACC issue

Post by bawjkt »

Just an update -

I turned ACC back on a week ago and it is running without event.

So, I have not seen this condition reoccur.

Call it a one-time event with an unknown cause.

Grey
Posts: 8
Joined: Wed Dec 12, 2012 7:16 pm

Re: ACC issue

Post by Grey »

Perhaps your provider changed something.

Latencies over 400ms though is quite easy to believe. Some modems have huge buffers that can cause delays of a couple of seconds.

Here is a link that will stress test your link Netalyzer. Tool was created for bufferbloat research and it does a great job of showing if you have an issue on your link, up and down both.

raz
Posts: 41
Joined: Sun Feb 21, 2010 11:12 am

Re: ACC issue

Post by raz »

Grey, thanks for the link!

Super useful tool!

pbix
Developer
Posts: 1371
Joined: Fri Aug 21, 2009 5:09 pm

Re: ACC issue

Post by pbix »

This type of problem could be caused by congestion in the uplink.

ACC only reduces the downlink bandwidth in response to long RTTs. If the long RTT is caused by the uplink there would be no improvement.

Not sure how uplink congestion might occur at a bar but it is possible I suppose. If you want to investigate the next time it happens try reducing the bandwidth of the uplink on the the QoS uplink page by 30%.
Linksys WRT1900ACv2
Netgear WNDR3700v2
TP Link 1043ND v3
TP-Link TL-WDR3600 v1
Buffalo WZR-HP-G300NH2
WRT54G-TM

Maniac
Posts: 70
Joined: Wed Sep 02, 2009 7:52 pm

Re: ACC issue

Post by Maniac »

Hi, I'm having this same issue - ACC is (correctly) limiting my link limit when my ping is going to 400+ ms (something seems to be wrong with my ISP). However, the problem resolves itself shortly, but ACC VERY slowly ups the link limit - in about 3 kbps increments. This means it takes over 10 minutes to get to my cap of 3 mbps from the 700 kbps it throttles to.

Can ACC be adjusted to increase the link limit more aggressively when the ping settles back to normal response times? This would fix the issue I'm having when ACC throttles my bandwidth. (Regardless, I need to find out why I'm randomly getting these high ping times).

Thanks!

pbix
Developer
Posts: 1371
Joined: Fri Aug 21, 2009 5:09 pm

Re: ACC issue

Post by pbix »

There are no user adjustments available for the control loop of the ACC. However, I have recently made some changes in source code (no in the repo) which should help. The ramp rate has been increased and when switching from RTTMODE<->ACTIVE MODE the ACC remembers its previous limits and starts the control from there. This should lead to faster settling on the final value.

To get these latest changes you will have to build Gargoyle yourself or wait until the next release.

There is a limit to how fast the ACC can change the limit or instability will be introduced in the control.
Linksys WRT1900ACv2
Netgear WNDR3700v2
TP Link 1043ND v3
TP-Link TL-WDR3600 v1
Buffalo WZR-HP-G300NH2
WRT54G-TM

Maniac
Posts: 70
Joined: Wed Sep 02, 2009 7:52 pm

Re: ACC issue

Post by Maniac »

OK, thanks. I'll let you know how it turns out when the next version of Gargoyle comes out.

Post Reply