Ping jitter to detect issue

My setup:

x86 OpenWRT router + Access Point with vendor firmware
I have 1Gbps ISP connection.

I am noticing that sometimes video streaming stutters.
I am using 802.11ac and sitting right next to the AP.

During problem I noticed that ping jitter to google from my laptop is high.
But ping out from the openwrt router is fine.

This is telling the AP lands into issue sometimes.
Recovers after a restart of the AP.

This got me thinking.

Is a continuous measure of ping jitter a good indicator of network health?
If I were to implement this, how would I got about this?
Will it be overload the router?
I have option to upload the raw ping data to a server.
Or calculate jitter locally and then upload final result.
Is there any app already available?

or write a shell script and add it to scheduled tasks.

1 Like

collectd perhaps?

1 Like

What does show? What is the bufferbloat grade it displays?

It may not be a problem of bufferbloat, but if there's latency when there's traffic, that could definitely add to the problem.

1 Like

It wasn't the issue on WAN, as ping out from within the router itself was fine.

Am sure it was the wifi access point.

yet to have a complete look at this.

However, does a sample of ping data contain valuable info.. wanted opinion on that if someone has tried this.

I mean, say in past 10 mins window I have 60 samples of ping results; 1 ping every 10s to the user's device from the openwrt router.
Can the jitter value give me a fair idea of internet experience for the user?

yes, ping and jitter give good indication of connection quality.
no, it will definitively not overload anything. google will not care about it either if you send 100's of pings per second because some dumbfuck youtube video will cost em faaaaaar more.

but ping and jitter may not be directly indicative of "browsing speed". ymmv

1 Like

But can they be directly indicative of wireless link quality?

To be clear I want to ping the wireless client from within the router.

Also I want to only take samples by pinging every 10s. Not every second which is the default.
So that the amount of data to be analysed/plotted/uploaded is manageable.

I think the challenge would be to find what value of jitter is problematic. I have seen good internet experience even while having very high ping jitter, whereas other times less jitter caused problem. I don't know if there is direct correlation.

but the dominant factors will probably be channel utilisation and power saving states.
10s is quite a long interval, so it will not really be telling much.


I need to read up on what they are.

Want to know why you think so.
I mean, it's just down-sample of the whole data, should tell the trend.

but i want to stop discouraging you here :slight_smile:
you will see some trends over a big enough measurement window.

1 Like

I calculated ping and standard deviation (jitter) and plotted in google charts.

This is how it looks.

Not yet sure what to make of it.

As per cisco jitter should be within 30ms for voip, so I am certainly exceeding that.