grrrr latency back to old tricks.

[#1]

We found 3 problems which were causing the latency and packet loss problems, 2 on BT's side and 1 on our side. The one our side was caused by a problem with the titanium queue on THN-AG2, it was partly fixed a few weeks ago and then fully fixed with the reboot of THN-AG2 a week or so later. The BT side problems were caused by them first not setting the policer to the right value and then after correcting that not setting the shaper correctly. Both those should now be fixed and we've not seen anything up since. Be interesting to see if anyone else is seeing it.

[#2]

Sorry, Dave
Fairly sure it's not fixed - if anything it's actually got worse in the timeframe you've mentioned.
Since about 2 or 3 weeks ago I've seen my average ping in World of Warcraft go up from ~100ms to ~200ms which is close to being unplayable.  I need my warcraft fixed!

[#3]

Ok, to add to my fairly sparse post...
I've done a series of tests and found some interesting results...
Test one - ping -t 212.159.13.49
This results in a stream of 30ms replies that are pretty much stable +- 3ms until something goes wrong and the ping will jump all over the place for 10-40 secs: 40ms, 49ms, 84ms, 54ms, 68ms, 74ms 77ms, 79ms, 45ms etc until all of a sudden it will be back at 30ms and stable again.  There is no pattern to the times between "spasms".
Test two - ping -t 212.159.13.49 Performed from a pc not on ADSL2+ (actually BT Business Broadband)
The ping to the server is 100% stable.  Running test 1 and test 2 side by side show that while test 1 is spiking test 2 is 100% stable.
Test three - pinging a server over vpn (testing vpn priority)
The ping is 100% stable.  Running side by side with test 1 shows that even when test 1 spikes test 3 is stable.
Test four - tracert to Blizzard's servers
Continuous running of traces show that for the majority of the time there will be no delays and 3x 48ms pings at the endpoint, occassionally it will catch delays of up to 200ms at some hops but these are rare and never all 3 readings for a given hop.
Test five - ping -t 212.159.13.49 Performed on 2 computers connected to the same PlusNet ADSL2+ router
Here's the most interesting test.  Both PCs start off stable at 31ms each.  They both spiked at the same time, but one PC dropped back to 31ms way before the other and are now displaying completely independent behaviour, with one PC showing spikes when the other is stable and vice versa.
What is definitely clear from these tests is that something is not quite right - but I'm baffled as to what that something is.

[#4]

Final test - Pingplotter
Note the Y-axis scale differs for each hop - vl604.thn-gw1.plus.net is spiking to 327ms 

[#5]

Hi Puffin,
I'm sorry that you're still having problems with your latency.
I'll have a chat with Dave when he gets back into the office tomorrow,

[#6]

puffin what happens to your pings over the time period of say 23.50- 00.20 ?

[#7]

I'll check later!
In the meantime, here's an interesting graph of activity that started at 16:42 today (before that pings were steady at 30-35ms).

[#8]

are those over wireless?
if so you might want to disable the WZC service of windows which if you are using wireless would stop that regualar spike every minute in your graph
http://www.ifelix.co.uk/tech/2000.html for xp
http://www.martin-majowski.de/wlanoptimizer/ for vista a little app that does it for you
my pings at 5.30pm
Ping statistics for 212.159.13.49:
    Packets: Sent = 132, Received = 132, Lost = 0 (0% loss),
Approximate round trip times in milli-seconds:
    Minimum = 15ms, Maximum = 26ms, Average = 16ms

[#9]

Nope, wired all the way.

[#10]

Ok, so my connection just went crazy at midnight.  It settled down a bit at 00:15, but there's substantial packet loss on all hops and average latency is up by about 20ms on each hop, too.

[#11]

I've been PingPlotting for 48 hours now and it's clear that there are two issues:
The first is a recurring issue that happens approximately every 90-100 mins. First pings to 212.159.2.124 (vl55.ptn-lb01.plus.net) become erratic, then 4 minutes later 212.159.13.49 (pth-cdns01.plus.net) becomes erratic, too.  The erratic pings last for 6 minutes for each server then pings return to normal.
The second occurs at 00:12 ever night and lasts for 15 mins.  Unlike the other issue, the pings to all hops become erratic at the same time and are exactly in sync.  After approximately 15 mins the pings return to normal and all is well.
Any chance we can get some PlusNet follow up on this behaviour?

[#12]

yeah we know about the midnight problem but thanks for picking up on the the nearly on the hour behaviour up.
does this even happen say 2pm till 9pmish just wondering as before anything was done to sort the pings when ever rate limits where increased the pings would spike for 5 mins after rate increase and ask if it happens between those time periods as this when limits capped down not up.

[#13]

Yes it's literally every 90-100 mins (not sure of the exact period but it's definitely happening throughout the entire day).

[#14]

Quote from: Jameseh

Hi Puffin, I'm sorry that you're still having problems with your latency. I'll have a chat with Dave when he gets back into the office tomorrow,

James/Dave:  It's been a week now.
Can you confirm you've seen the graphs posted in this thread and are working on providing an explanation and fix in the near future?

[#15]

Sorry Puffin - I hadn't intended to leave this dormant.
We're aware of latency concerns and will be working to resolve.