EE latency problem - advice sought

Hi

I'm new to this forum (and broadband technical stuff generally) so please forgive any naive questions.

I have a problem with my EE ADSL broadband which I think is one of latency rather than speed, experienced as an intermittent slowness in reaching (rather than downloading) web pages. The problem is the same whether the pages are information-heavy (e.g. video) or information light (e.g. my email in-box). Typically, it can take more than 30 seconds to get through to a page whose servers I know to be located in London (i.e. my university's email server), and sometimes I get timed out altogether ('Safari was unable to open the page because the server stopped responding').

The thinkbroadband speed test does not capture this problem - it gives consistent speeds of about 9 Mb/s down and 0.9 Mb/s up and a ping time of about 30 ms. However, with other speed tests (e.g. broadbandspeedchecker) I have been able to record ping times in whole seconds - enough to reach a server on the moon, by my calculations!

There are two tests which have shown something correlating with my experience:

(i) The broadband quality monitor shows frequent flares in latency, e.g. yesterday:

My Broadband Ping

The timing (7am - 9am) would seem to suggest congestion as the cause, but similar episodes have occurred in the past at all times of the day, including 2 am at night.

(ii) I made over 20 traceroute maps to the usual sites (bbc, google etc). All of them got timed out at step 2 and most (17) of them also at step 3 eg. :

traceroute: Warning: www.bbc.co.uk has multiple addresses; using 212.58.246.92
traceroute to www.bbc.net.uk (212.58.246.92), 64 hops max, 72 byte packets
1 brightbox.ee (192.168.1.1) 1.021 ms 0.625 ms 0.494 ms
2 * * *
3 * * *
4 213.121.98.136 (213.121.98.136) 7.320 ms 7.640 ms 7.106 ms
5 87.237.20.132 (87.237.20.132) 8.066 ms 8.086 ms 15.801 ms
6 bluestream-gw-b.thdo.bbc.co.uk (212.58.239.54) 7.622 ms 8.524 ms 8.337 ms
7 * * *
8 132.185.254.93 (132.185.254.93) 10.813 ms 11.096 ms 10.816 ms
9 132.185.255.164 (132.185.255.164) 14.270 ms 13.836 ms 14.798 ms
10 bbc-vip013.cwwtf.bbc.co.uk (212.58.246.92) 13.387 ms 13.507 ms 13.272 ms

I interpret this as repeated failed attempts to access a gateway server belonging to or leased by EE (213.121.98.136 is listed as a BT server, 87.237.20.132 as Orange).

I also notice that although all the route times add up to less than 200ms the traceroute map took about 2 minutes to complete (is this normal?).

Needless to say, I have raised the issue many times with EE to no avail. They have not been able to find anything wrong with my setup, either remotely or when an engineer visited my house. Perhaps they consider this level of service to be acceptable, but I need to do something about it. I am employed by a university but often need to work from home, using internet to communicate with colleagues in real time.

I have two options:

(i) EE suggest upgrading to their basic FTTC service for about £30/month. But I can see no technical reason why that would solve a problem caused by congestion on their servers (unless they actually prioritise traffic for clients on the more expensive fibre tariffs for commercial reasons, which is possible).

(ii) Move to an ADSL service with one of the smaller, more expensive ISPs with a better reputation for customer service (Zen Internet?) for about the same price.

Given that current speeds are perfectly adequate for my purposes (I don't do streaming or gaming), and have lost confidence in EE, I'm inlcined toward the latter, but would appreciate any suggestions.

Sorry for the long post, but I thought best to put up all relevant information first. Any comments or suggestions would be appreciated.

David

That does look more like congestion than anything else to me.

The rows of three asterisks in the list of steps is the easy bit. Those occur for (mainly) two reasons.

First, tracert just pings each hop/router of the route in turn, several times, and that router can simply be set not to respond to pings in the same way as you can set yours.

Second, the important job of a router is to forward the packets in each direction as efficiently as possible. Many of them will prioritise real data when they are getting busy, and simply ignore pings at such times. You can sometimes see the effect of that when a line will have one or two stars, and a reading in the other position(s).

The pure yellow spikes in your BQM are quite normal if you are using the link a fair bit. The surge between 6am and 8am suggests you were doing something a bit heavier, either uploading or downloading. To me the fact there is some red at the top, meaning dropped ping packets, could mean you were uploading. The heavy surge at 8:30'ish to just after 9 maybe a big download or streaming? That is fairly common.

Slow page and email loads on the other hand, which is your main complaint, is almost certainly congestion, though maybe at different points. For example a university mail server itself may be under pressure, and links to other web sites from EE could be an EE problem. It's hard to know.

Just a suggestion re your BQM, it might be a good idea to include the "Share live graph" link in your sig, like mine, so we get a better view of how things vary .

I could be wrong, but this looks to me like more of a DNS problem.

If speedtests are showing that you can download and upload at the speeds expected then I doubt it's congestion (ignore broadbandspeedchecker as it's not always very reliable).

You say it's the loading of pages that seems to be the issue and that would point to a DNS issue, have you tried changing your DNS settings to another DNS provider (like google for instance)?

7am to 9am is a very odd time for congestion, usually it is at night.

A link to a speed test result done at https://www.thinkbroadband.com/speedtest in particular if this is very slow DNS lookups I would expect to see a delay in the download testing when you look at the graphs at the end.

That traceroute took so long is pretty irrelevant to your problem

The latency we can see looks like nothing is particularly wrong, so DNS lookup times seems likely as others have suggested.

Dear Roberto

Thanks for your comments.

Here is a link for the live BQM:

My Broadband Ping

As you can see, there was another episode last night at 1 a.m. - the only activity I can find on the system log at that time is 'Time is synced to NTP server' which I guess is some kind of routine automatic communication between my router and a remote server to check time synchronisation, but surely not enough data transmitted to cause packet loss, significant increase in average latency etc. ???

In other words, the pattern of spikes does not seem to be related to my use of the link, which is as I have said fairly light on data (no streaming, no gaming).

Anyway, I am learning something, which is the main purpose for me of coming here ...

Dear Robert

Thanks for your comments and suggestion.

If I understand you correctly, this means that the default DNS server used by EE may be congested, and by changing settings on the router (it is an EE 'Brightbox' model) I can use another preferred DNS server?

Would this be found under:

Broadband settings>DNS IP address>Manual ??

It is currently set to 'Obtain from ISP', and there are two boxes to enter named DNS IP's ('primary' and 'secondary') which are currently empty.

And what is the address for the google DNS server?

With regard to your theory, I have just tried reaching some web pages using the numerical IP address, and it definitely seemed to speed things up. But I will have to keep testing over a longer time period to be sure.

You can manually change it to something like google which is
8.8.8.8
8.8.4.4

Dear Andrew

Thanks for your reply.

Here is a link to a speed test just completed (at 12.50 pm):



As usual, it doesn't seem (to me) to show anything amiss.

But I don't seem to be able to post a link which shows the graphs, which is presumably what you wanted o look at?

I am interested in your comment 'the traceroute took so long as to be irrelevant'. Do you mean that there was something abnormal about the traceroute map, but that it might indicate a different problem besides latency? Or that all traceroute maps are inherently slow, and therefore irrelevant?

In any case, I didn't quite understand that bit.

The link is enough, the fairly classic flow ramp up of speeds to the maximum, should be a much sharper rise at the start and is characteristic of EE, which will make web pages feel slower

The DNS does not look massively impacted.

On the traceroute no nothing abnormal, not all devices respond to pings and many have slow name resolution hence why it takes a while to perform.

Solution might be to switch providers.

What the cause of this classic EE shape is is an unknown, maybe the Brightbox itself.

But I don't seem to be able to post a link which shows the graphs

Clicking the graphic you posted shows the underlying graphs.