BB performance very poor in last few days

Since sometime over the weekend my BB performance has become very variable. Previously I usually got reasonable response times to most web sites, speed tests typical show 8-9 Mbps D/L, 0.9 Mbps upload, enough for my needs. Now web site response is very laggy, especially noticeable on, for example, BBC Weather site. Also on https connections.

TBB Speed tests this week have shown D/L figures 25% or less of those (2 Mbps on a good day); U/L broadly ok. Just run the BTW speed test and it shows 1.12 Mbps D/L; 0.73 Mbps U/L; ping latency (ms) 0.00 (!?)

I've also tried things on my laptop, normally run on my desktop, and also my wife's iPad. All are giving much the same results, and also much the same web site lag.

Router stats are
CS (kbps) 8659 / 1127
LA (db) 42.0 / 24.5
NM (db) 8.9 / 5.7

The router told me there was new firmware so I've installed that , no difference.
I tried an old but faithful Netgear non-wireless router, no difference.
I've set the PC's network connection to use Google's DNS, no difference.
I've set the router to use Google's DNS servers, no difference.

Web transactions are very slow, for example to get to Paypal's payment page a few minutes ago took several minutes.

A download of a new version of Ccleaner last night stopped at around 0.9 Mbytes, and no more, even after 10 minutes.

I keep an eye on Win 10's Task Manager Ethernet monitor and frequently see download of 0 for 30 seconds or more even when waiting for a page or download.

Occasionally it's all fine for a few moments, but only for a few moments.

Both my desktop and laptop are running Win 10.
ISP is Eclipse.
Router is Netgear DGN2200v4 running V1.0.0.86_1.0.86 firmware.

I haven't yet spoken to my ISP as i'm really not sure what to tell them.

Any thoughts?

Link to a speed test there may be some clues in the graphs, e.g. as well as lower figures the lines are more wobbly than usual

Thanks.

This is one I've just taken. Interesting,perhaps, that it spent 20 secs or so with the spinning disc before it even started, and Task Manager was showing zero download traffic for that period until it suddenly burst into life.

Which ISP please, Tony?

Eclipse, as said in OP.

D'oh!

I should have mentioned that I took my laptop up to our club last night and it connected to the WiFi there. A tbb speed there test showed around 30 Mbps. So it's no a PC issue.

It's surely something to do with the router or outward of there.

Look at latency figure 3000ms (3 seconds) that explains why you have the experience you describe.

The spinning disc at start is the establishing the base line idle latency, so the fact it took so long indicates not a maths error but really slow lookups.

All the latency checks are

2013,27709,117,37,37,37,37,38,38,37

In short very slow and then once something has woken up all is well. This need to wake up is not normal, it should not be DNS since you've accessed the domain already that the latency calls are made to

Given PC is fine elsewhere suggests router/ISP issue

I had a conversation with Eclipse Tech.Support last Wednesday and they made some Exchange changes. No difference. He asked me to try a different cable from router to wall socket and to use the test socket. No difference.

The performance is still poor, though the one speedtest result I posted that had a 3000ms latency seems to have been an outlier in its latency result. Others are showing latency in about the 20-50ms range.

I can't detect any line noise.

I've now raised another ticket so we'll see what happens over the coming days, after the Bank Holiday.

The router stats continue to give me figures that are historically normal and not ones that would give the poor performance that I'm currently getting. Even downloading PDF files from my church's website (that I manage) shows great performance variations and stalling.

It seems to point to external causes, we'll wait and see.

Edited by cheshire_man (Mon 29-May-17 09:30:59)

Note how those high latencies are at the start of the train, and this matches the description you give for pages loading

i.e. once the router/ISP wakes up things are ok

Yes Andrew, the latency doesn't appear to be an issue now. But the download performance is still only around 30% of what I was getting before, and getting for years.

Something has changed, gone faulty, in the last 10-12 days.

I've reset my cabling etc to its normal, normal socket, original 'modem' cable. No significant change in performance. Router statistics sill fine.

One thing I've noticed though on speed tests is that the x6 figures are significantly higher than the single thread, perhaps 5-7 times.

This morning's quartet:

Test 1 @ 6:29 before any changes - latency 51ms


Test 2 @ 6:48 after changing modem cable and using test socket - latency 59ms


Test 3 @ 6:51 same set up as Test 2 - latency 28ms


Test 4 @ 9:24 after reverting to normal socket and original modem cable - latency 99ms


Does that suggest congestion?

What else?

Edited by cheshire_man (Mon 29-May-17 09:45:26)

The differences do suggest congestion and if you look at the shape of your httpx6 results they are not nice and flat

e.g. https://www.thinkbroadband.com/speedtest/14960481988...

I had a call from my ISP Tech.Support this evening. After 30 minutes of conversation and him checking all sorts of things, he decided to book an Openreach engineer. He was surprised to find the request bypassed the BT diagnostic team and straight to an SFI engineer booking, due sometime tomorrow afternoon.

During the call he asked me to do a Ping x 100 to Google's DNS - 8,8,8,8. Both then and a couple I've run since show that 17-20% are timing out, those that don't get a response in 14-15ms.

We'll what happens tomorrow.

An SFI generally only going to worry about the state of the DSL line, your issue is more likely to be somewhere beyond the exchange and before Internet, i.e. ISP or wholesale element of backhaul

The engineer came on Wednesday. His first question was "Who fitted the wall box?" "BT" I replied, it was an ADSL 1.0 ftted perhaps 10 years ago while sorting out some damaged wiring. He changed it for a Master Socket 5C.

His tests using his equipment all showed a D/L of around 9.5 Mbps. But the BTW speed tests showed around 1.5-2.5. Changed the router, no difference, similarly the Ethernet cable. He used my modem cable for his equipment and got 9.5 Mbps again so not that cable. I pointed out that my wife's iPad and my laptop are both giving similar results to my desktop PC, and that the laptop got ~30 Mbps over WiFi at our club, which IMO rule out a PC problem.

But of course his speed testing was only back to the line card in the exchange. I suggested that as 'my' speed tests were all to the internet, i.e. beyond the exchange it suggested a more distant problem. He did agree that it could be a line card problem. I asked him to include that in his notes.

Oh, fortunately he was on a diet as when I offered him a cuppa and had to apologise as we don't have biscuits in the house, he declined

Got a call form the ISP this evening who had now seen the engineer's report and has requested a 'lift and shift', this has been accepted by BTOR and should be done within 48 hours, so we'll see what that does.

Good luck 🙂

Can you post up links to your Speedtests please? Not images.

I've been there and done that with lift and shifts etc. Don't expect too much.

If this is congestion will you be happy to switch provider?

URLs for the 4 speed tests early in the thread:

Test 1 @ 0529

Test 2 @ 0548

Test 3 @ 0551

Test 4 @ 0824

I presume those are what you were asking for.

As far as potentially switching provider, I'd be reluctant to. I've been with them since 2003 and have been very happy with their service and the way they've bent over backwards to help me with the (very) occasional issues I've had.

I'll give them absolutely every chance to get the issue resolved.

How interesting?

Noticed that pages that had been slow - banking especially - for the past 2 weeks or so were responding as of old. So did a speed test



Or in more detail

I wonder if the Lift and Shift has already been done?

As of this moment certainly looks to be back to it's normal self

I'll wait to see what the ISP says next week as to when the Life & Shift was done.

Edited by cheshire_man (Sat 03-Jun-17 16:58:27)

I'll wait to see what the ISP says next week as to when the Life & Shift was done.

It will be interesting to see what they say ....... I seriously doubt that a lift and shift has resolved anything.

Had you tried any trace routes whilst it was poor, and if so can you now compare those with some done now ? This might give detail as to where the change has occurred .

No, have no trace route data from when it was poor.

Just done 100 pings to 8.8.8.8. As mentioned earlier, when it was poor, some 15-20% were timing out; this time 0% timeouts, all times were 14-16ms.

Whatever's been done, and clearly something has been done, it's back to normal.

Yes, that it's fixed is the main thing.

Purely as a matter of interest, I assume a Life and Shift would, albeit briefly, cause the link to drop.

If that's the case, then as the router reports the WAN being up for 1 day 19 hours 5+ minutes, it suggests the change was done around 1230-1300 on Saturday.

At nearly 2 days that WAN up time is the longest I've seen for a couple of weeks.

Edited by cheshire_man (Mon 05-Jun-17 07:48:29)

I assume a Life and Shift would, albeit briefly, cause the link to drop.

Yes, and more than briefly too.