Severe Latency Issue

[#1]

Hi all,

I'm hoping someone can help me diagnose a problem that has appeared recently on my Plusnet FTTC connection. I've been with Plusnet for around five months and, until recently, the connection has been fine.

My normal download speed is only around 16–17Mbps because of my location, but that has always been the case and isn't really the problem. The issue is that web browsing has recently become very slow/intermittent, with pages sometimes hanging before loading.

I started testing latency to 1.1.1.1 and initially got some horrendous results. One 30-ping test averaged 728ms, with a maximum of 1018ms and some packet loss. Another averaged 897ms, reaching 1154ms.

Rebooting the Plusnet Hub Two initially restored the connection to around 12–15ms, but after roughly 10 minutes the very high latency returned.

I initially suspected my mesh Wi-Fi system. I therefore completely unplugged the mesh and connected directly to the Plusnet Hub. Ping to the Hub itself (192.168.1.254) is excellent: 1–4ms, 2ms average, 0% loss, so the local Wi-Fi connection to the Hub appears fine.

With the mesh completely disconnected, idle internet latency is also excellent at around 11–13ms. However, I've established that latency rises dramatically as soon as the internet connection is under load.

During a speed test, the download portion raises the ping from ~12ms to around 60–70ms. As soon as the upload test starts, it goes to 815ms, 1071ms and then request timeouts. I've also seen ordinary web browsing produce shorter spikes into the 200–470ms range.

This behaviour is definitely reproducible with the mesh system completely unplugged, so whatever is happening is not simply a mesh/Wi-Fi fault.

The Hub Two DSL statistics currently show:

Data rate: 3.085Mbps up / 18Mbps down
Maximum data rate: 3.085Mbps up / 19.392Mbps down
Noise margin: 6.0 / 7.3dB
Line attenuation: 11.2 / 27.1dB
Signal attenuation: 11.2 / 34.0dB

Speed tests normally give about 16–17Mbps download and somewhere between 0–2Mbps upload.

I realise that saturating a relatively small ~3Mbps upstream can cause bufferbloat, and the upload speed test itself is deliberately saturating the connection. What concerns me is that this has only become a noticeable problem recently after approximately five months of normal use, and even relatively ordinary activity can now sometimes cause substantial latency spikes.

One potentially relevant factor is that Openreach have recently been carrying out extensive FTTP installation work throughout the surrounding area. Our property is on a private lane and isn't currently getting FTTP, so we are still using the existing FTTC/copper service. I don't know whether work in nearby cabinets, chambers or ducts could have disturbed the existing copper line, but the timing made me wonder.

I'd particularly appreciate it if somebody could advise whether the 3.085Mbps upstream sync/maximum attainable rate looks reasonable for this line, and ideally whether Plusnet staff can check my historical upstream sync rates to see whether this has recently fallen.

I'm also interested in whether these results point more towards a line/DLM/profile problem, a Hub Two problem, severe bufferbloat that has always existed but is now being triggered more often, or possibly a physical copper-line issue following the recent Openreach work.

I've already submitted the issue to Plusnet, but I'd be grateful for any interpretation of the stats or suggestions for further tests. I'm happy to provide additional DSL information or run specific tests if useful.

Thanks.

[#2]

@silkster365 Your D/S  SNR  looks a little high for your reported attenuation.

Does the the event log give any clues - DSL  drops, or suchlike?

[#3]

Thanks John. I've attached/checked the WAN event log. The only explicit DSL Link Down / ERROR_NO_CARRIER entries I can see are from 27 Feb and 1 Mar; I don't appear to have had repeated DSL drops recently. Historical sync also seems quite consistent — upstream was generally 3200kbps in March, 3158 on 18 Jun, 3106 on 25 Jun, 3200 on 16 Jul, 2917 yesterday and 3085 this morning. Downstream has consistently been ~18000kbps. Downstream SNR was actually around 9.6–9.9dB in March and is 7.3dB now. Does that change your view of the SNR/attenuation figures?

[#4]

@silkster365 Thanks for that event log. It concerns me a little that there are no corresponding 'DSL Down' entries, but I'm no networking expert.

Maybe @Dan_the_Van  will have some thoughts?

[#5]

@jab1 

It concerns me a little that there are no corresponding 'DSL Down' entries

You only get the DSL Link down messages written to the event log if the router is powered on.

A restart or power off of the router does not record that event as the DSL Link wasn't down at that point.

Looking at the event log, which has had a category filtered applied so not All event are present.

@silkster365 

Yes having a slow upload speed isn't really helpful for the modern world of the internet, devices with cloud saves/backups taking up the upload bandwidth.

The mesh system you have does it have a router mode with QoS/SQM settings? You could set the Hub Two to 'bridge mode' which makes it a modem only device, then you'll be able to take advantage of router functions.

[#6]

---

@silkster365 wrote:

I'm hoping someone can help me diagnose a problem ...

---

Sigh !,  where to start ?

---

@silkster365 wrote:

.... a problem that has appeared recently on my Plusnet FTTC connection

---

Are you -

1. really still on FTTC with a landline phone ?
2. on FTTC but without a phone attached ?
3. or upgraded to SoGEA (FTTC with the phone service deactivated) ?

How is your Plusnet router plugged into the BT landline ?

1. Is it plugged in to the BT master socket, or a via BT extension socket ?
2. Does the master socket have a DSL filtered faceplate, or is it a single phone socket with a dangly external DSL filter ?
3. Does you house have ANY phone extension wiring ?
4. How long is the cable between the phone socket and your router ?

---

@silkster365 wrote:

I've been with Plusnet for around five months and, until recently, the connection has been fine.

My normal download speed is only around 16–17Mbps because of my location, but that has always been the case and isn't really the problem. The issue is that web browsing has recently become very slow/intermittent, with pages sometimes hanging before loading.

---

Have you recently changed your mobile phone, or added a new device to your network, that is now doing cloud data backups ?

Apple iPhones and iPads are notorious for causing this type of issue, as they are designed expect the connection to not bufferbloat, and will grab as much bandwidth as possible to give the illusion of better performance (compared to Android alternatives).

Do you have something on your network that otherwise does a lot of uploading, such as CCTV cameras, or video sharing ?.

---

@silkster365 wrote:

I started testing latency to 1.1.1.1 and initially got some horrendous results. One 30-ping test averaged 728ms, with a maximum of 1018ms and some packet loss. Another averaged 897ms, reaching 1154ms.

Rebooting the Plusnet Hub Two initially restored the connection to around 12–15ms, but after roughly 10 minutes the very high latency returned.

---

That is typical Plusnet Hub Two behaviour, on a slow upload speed connection like yours, when something on your network is uploading continuously.

---

@silkster365 wrote:

With the mesh completely disconnected, idle internet latency is also excellent at around 11–13ms. However, I've established that latency rises dramatically as soon as the internet connection is under load.

What is the exact model of mesh system you have ?

Are you -

1. Using the Plusnet hub as a modem+router, and the mesh acting as wireless access points
2. Have you got the Plusnet hub in 'bridged VDSL modem' mode, and have the mesh in 'router' mode ?

I'm thinking as the Plusnet hub two doesn't have any anti-bufferbloat settings, that if your mesh supports 'router mode', then many mesh systems do have anti-bufferbloat settings - so you might be able to solve your problem by reconfiguring your mesh if it has some form of SQM, QoS, rate limiting, CAKE, fq_codel, etc.

---

@silkster365 wrote:

The Hub Two DSL statistics currently show:

Data rate: 3.085Mbps up / 18Mbps down
Maximum data rate: 3.085Mbps up / 19.392Mbps down
Noise margin: 6.0 / 7.3dB
Line attenuation: 11.2 / 27.1dB
Signal attenuation: 11.2 / 34.0dB

Speed tests normally give about 16–17Mbps download and somewhere between 0–2Mbps upload.

Those stats look OK for a long line.

Your FIRST move is to do EVERYTHING you can to optimise your home's internal phone wiring, get the router as close as possible to the master socket, and if necessary have the router's DSL port connected to the hidden 'TEST' socket behind the master socket faceplate.  Minimising any noise or interference on your broadband connection, should improve your sync speeds, and reduce the bufferbloat.

---

@silkster365 wrote:

I realise that saturating a relatively small ~3Mbps upstream can cause bufferbloat, and the upload speed test itself is deliberately saturating the connection.

---

Yes, the Plusnet Hub Two often has this problem at those speeds.

Do you happen to have any other broadband routers (from previous broadband contracts) available to experiment with ?

---

@silkster365 wrote:

One potentially relevant factor is that Openreach have recently been carrying out extensive FTTP installation work throughout the surrounding area. ... ... I don't know whether work in nearby cabinets, chambers or ducts could have disturbed the existing copper line, but the timing made me wonder.

Maybe, but unlikely.  If anything if the other customers in the surrounding area are moving from VDSL to full fibre, then the electrical crosstalk on your line should reduce (as their lines go quiet), and therefore your line should become less noisy and potentially faster.

---

@silkster365 wrote:

Our property is on a private lane and isn't currently getting FTTP, so we are still using the existing FTTC/copper service.

Have you checked whether you are in the connection area of any AltNet (non-Openreach) full fibre providers ?

If you let us know your postcode, then I could do a search for you to see if you have better options.

---

@silkster365 wrote:

I'd particularly appreciate it if somebody could advise whether the 3.085Mbps upstream sync/maximum attainable rate looks reasonable for this line

---

If you go to the BT Broadband Availability Checker , then enter your phone number or one of the other options, and get to the results table and explanatory text.  Take a screenshot from the top of the purple table to the bottom of the screen but EXCLUDING your phone number or address at the top.  Then post that image back to this forum topic.

---

@silkster365 wrote:

I'm also interested in whether these results point more towards a line/DLM/profile problem, a Hub Two problem, severe bufferbloat that has always existed but is now being triggered more often, or possibly a physical copper-line issue following the recent Openreach work.

Based on the information so far, something recently on your network has increased the background upload traffic, and has tipped the Plusnet Hub Two (which is particularly susceptible to this) to bufferbloat.

---

@silkster365 wrote:

I've already submitted the issue to Plusnet,

---

Well good luck with that,  I've never seen any response from Plusnet front line support that indicates they understand what bufferbloat is or how to cure it,  particularly since they removed traffic prioritisation in June 2017.

Plusnet Removes Broadband Internet Traffic Prioritisation 

Going forwards -

1. Upgrade to full fibre if an AltNet network is available to you
2. otherwise, optimise your wiring around the BT master socket to give your VDSL modem the best connection
3. check whether your mesh can be configured to rate limit it's WAN connection to overcome bufferbloat
4. if there is still a problem, consider buying a 'gaming' router that supports some form of anti-bufferbloat SQM, QoS

[#7]

Thanks, this is useful.

I've run the BT Wholesale checker and attached the screenshot. It shows:

VDSL Range A: 20–30.8Mbps down / 3.8–5.6Mbps up
VDSL Range B: 17.8–30.2Mbps down / 3.6–5.5Mbps up

My Hub is currently synced at 18Mbps down / 3.085Mbps up, so downstream is close to the bottom of the impacted range and upstream is actually below the 3.6Mbps bottom estimate.

On the wiring side, the Hub Two is immediately next to the BT master socket with only the short DSL lead between them. The master socket itself was replaced with a brand new one by an engineer around 12 months ago. There isn't a long internal cable run between the master socket and router.

I'll check whether there is any remaining extension wiring connected to the master socket and can test from the hidden test socket if that would still be worthwhile.

The mesh is a TP-Link Deco AC1300 system. The Plusnet Hub Two is currently operating as modem/router and the Deco is explicitly set to Access Point mode, so QoS/routing on the Deco is currently disabled. I have not put the Hub into bridge mode.

I don't have CCTV cameras or any other device that I know to be continually uploading. Nothing significant has deliberately been added to the network recently, although I'll investigate cloud backups/sync in case something has changed without me realising.

One other potentially useful point is that the ~3Mbps upstream isn't completely new. Looking through the WAN event log, historical upstream sync has generally been around 3.1–3.2Mbps for several months. What is new is the severe latency/browsing problem.

FTTP unfortunately isn't a realistic option at present. We live on an unadopted/private road and the permissions required for Openreach to bring FTTP along it cannot currently be obtained. The BT checker also confirms FTTP is unavailable at the property.

Given the Wholesale estimate has a 3.6Mbps minimum even for an impacted VDSL line, does the current 3.085Mbps upstream suggest there is still some value in investigating the DSL side before I start reconfiguring the Deco/Hub for rate limiting?

If we do conclude it's predominantly Hub Two bufferbloat, I'd also be interested in what upstream cap you'd recommend starting with on this ~3Mbps connection.