Little peak time latency humps

[#1]

But if less than 0.5% of pings are higher than normal - does that actually affect or cause problems with anything?
Anyone on wireless is going to see a far greater % of higher pings within their internal network.

[#2]

That's very debatable.  We don't know, but it certainly doesn't look good on those graphs!

[#3]

Here is my graph from yesterday Sunday 18th August. Connected to pcl-ag01.

On and off using the phone throughout the day for a mix of youtube and gaming and browsing. Gaming between 9pm and 11pm, was playing BO2 on the xbox. was mostly alright, odd slight laggy game depending on host, couple of bad spikes as well.
Any comments/thoughts?

[#4]

@ Mathias
Do you have something running 24/7 on your network? Your line seems to be active all day/night long.

[#5]

Quote from: AndyH

does that actually affect or cause problems with anything?

You keep saying that, Andy.
And at the same time contribute to the thread... And apparently throw a thousand+ pings at the BBC for no reason, therefore.
Conflicted?
I can't say I'm personally deeply troubled by 2 pings in 100 taking 17ms rather than 3. I wouldn't complain about it personally. But it might point to something else. Something else which does become significant at scale or under stress.
And who wouldn't want to know why? The terminally blinkered? Certainly not a Star Trek fan!
Why is your car making that funny noise? Well, who cares - it still works. Just ignore it until it breaks unexpectedly, at the least convenient moment, after which it's too late to do any testing or scoping of the problem (which wasn't a problem, of course, until it became one when the wheel bearing seized up at 70Mph in the outside lane!)
It doesn't make sense to make a fuss; but it also doesn't make sense not to investigate.
The fact that wireless adds other factors and makes things even worse isn't relevant, because to test something you have to remove variables, not add them.
Traffic shaping is very complex, not necessarily intuitive and there can be unexpected (and unintended) effects which only show up under marginal conditions. It's easy to get it slightly wrong, and it's better to optimize the setup when things are still working well so that the majority of customers who don't visit the forum don't ever start noticing a mistake and shouting about it.
Furthermore, the theory about what may be causing the effect needs to be tested from different angles because two different causes can easily result in the same effect. If the wrong solution is applied on the basis of an inadequately tested theory, it may appear to work for a time... until it doesn't. Then you have a new starting point to work from which may mask the original problem.
I think the fact that PlusNet are pro-actively acting on information available to them in order to provide the best possible service (and look good on graphs!) is a good thing. Their willingness to revise their approach on the basis of additional information as it becomes available, because they have sought it out by starting threads like this is even better. Certainly marks them out from most other ISPs.

[#6]

Quote from: SuperZoom

You keep saying that, Andy. And at the same time contribute to the thread... And apparently throw a thousand+ pings at the BBC for no reason, therefore. Conflicted?

Not really.
The 'ping spike issue' has been going on for many months. There's another very long thread somewhere where we looking at the issue back in Jan - Apr. But at that time, a serious issue developed where people were seeing massive speed slow downs on certain gateways during peak times (I could go from 86Mb/s down to 5Mb/s by changing gateways) and the speed/ping issues were initially linked. A lot of us gave input and help to dave/kelly + rest of PN staff to help them figure on what the problem was and it took a little while, but eventually the speed issue was diagnosed and fixed. The ping spike issue remained, but everyone's performance problems were fixed.
The ping spikes are not unique to PN - you only have to look at some BT Infinity, VM and Sky TBB graphs to see it's a pretty widespread problem. Anyone with an Asus N66U router *should* have the same problem due to a bug -- CPU overloaded?
I just wonder what people's expectations are and what can be realistically provided.

[#7]

You see, it isn't about what the latency humps are giving us, but what they're telling us.

[#8]

Quote from: SuperZoom

but what they're telling us.

That the network is busy?

[#9]

Actually, it's telling us there isn't enough titanium

[#10]

Changes have been rolled out, it's across the whole network and takes effect immediately so no need to disconnect/reconnect or anything. So long as there aren't any problems we'll leave it in place until Thursday and then backout the change. If we see improvements then we'll plan in when to do this permanently.

[#11]

Quote from: Kelly

Actually, it's telling us there isn't enough titanium

I can help - just put all my traffic in titanium at the network level!!!   
Oh. Actually, it would be inconsequential. I'll have to think of some other reason to justify doing that   

[#12]

Quote from: AndyH

@ Mathias Do you have something running 24/7 on your network? Your line seems to be active all day/night long.

No, i think that is just the effect of the router (as per the explanation of the graphs on the TB website). Using the Thomson 587v7 router. Still shows significant bumps in yellow ping though when the internet is in use (although distorted by the regular peaked nature).

[#13]

Quote from: dave

Changes have been rolled out [...]. If we see improvements then we'll plan in when to do this permanently.

Well, I don't know whether it's the same thing you're trying to cure or some entirely different phenomenon that Andy and I are seeing, but no change here in that I'm still getting 1 or 2 elevated pings per 100 (at 1 per second) using both icmp and tcp:
ICMP
Min:  4.2    /    Avg:  4.9    /    Max:  19.2
TCP
Min:  4.3    /    Avg:  5.3    /    Max:  19.5
That's using www.sky.com as a target for a bit of variety, tested just now over a run of 100 'pings' each on ptn-ag01.
Four duplicate SYN-ACK packets with the TCP.

[#14]

SuperZoom
sudo hping3 --icmp -c 100 www.bbc.co.uk
--- www.bbc.co.uk hping statistic ---
100 packets transmitted, 100 packets received, 0% packet loss
round-trip min/avg/max = 16.7/17.1/18.6 ms

sudo hping3 -S -p 80 -c 100 www.bbc.co.uk
--- www.bbc.co.uk hping statistic ---
100 packets transmitted, 100 packets received, 0% packet loss
round-trip min/avg/max = 17.9/18.4/19.1 ms
Do i press a special button to get a return of 3?
do i need to purge this software by the way ???? is it EVIL 

[#15]

  Yeah, I have a big red one on my keyboard - marked SuperZoom, of course!    
Ask Openreach to send you one!
It can be pretty evil software, actually, if misused. Go easy on the purging, though  
Your figures look pretty consistent there. I wonder whether there's something significant about the max figure, then, rather than the differential?...
Maybe see what it looks like during the peak.