Fibre Broadband Packet Loss Mystery

[the54thvoid]

Hi,

So, I am in UK using Virgin Media's broadband. I'm getting 500+ down, 35+ up. Fibre to wall, co-ax runs insode house and splits to TV and Router. Sounds okay so far.

Problem is (I have two), I'm plagued with drop outs which I know are from Virgin's side because of chekcing website 'down'detectors spiking when I have my disconnects.

Second issue I noticed gaming with stats up, was packet loss. I downloaded winMTR and ran some tests. I traced to TPU, and also my own connection. I did this on wired, and wifi (the wifi was done on a new laptop). Here are my MTR snapshots:

To TPU via wired PC (this is with 0.2 response time, same for all)

This is laptop wi-fi to TPU:

And laptop wi-fi to router

I've had packet loss before but never used MTR to check it (did so on advice this time).

Curveball - I ran the router diagnostics and did a ping test to TPU (10 pings), and it recorded 110ms with no packet loss on all ten.

Any ideas what would cause this, or how to trouble shoot? I have zero networking skills.

Edit:

Just connected to a 51ms server on AirVPN. Got this:

Still packet loss but way lower. Does this imply a problem with Virgin's line routing?

 

[Frick]

If the laptop can't traceroute to the router over wifi I'd say the router is the problem, and looking at the desktop traceroutes yeah.

 

[the54thvoid]

I just ran the laptop wifi again:

 

[tabascosauz]

Doesn't true fiber (fiber to house, into ONT, to ethernet out) always have symmetry of download/upload speeds? Only that allows you to exclusively use your own router hardware directly on ONT. Which means the router has to be some kind of modem+router combination, and even if you use your own router it has to go through their modem first. Crappy hardware is crappy. Bad experience with Virgin when I was last there.

How's the traceroute results in cmd look?

 

[the54thvoid]

Like this:

Ran this too (I'm googling as I go):

 

[Solaris17]

Can yall just not have network problems while im busy? I'm just trying to enjoy this scotch.

Finally someone that runs an MTR thank you.

So the MTR doesnt look good cheif. I dont have a ton of time at this isntant but next test is to ping the gateway. (your router)

Start small. ping 192.168.0.1 for a bit and lets see the loss.

 

[the54thvoid]

Done that - It's above - the single line MTR's.

Unless you mean cmd... I'll do it:

No loss.

I also re-ran winMTR and it still has low packet loss. Re-ran cmd trace and it has zero..... ???

One final 100 packets via cmd and it had 0% loss.

Win MTR - still has loss, at only 1% at 100 packets.

 

[Frick]

Faulty hardware can be like that. Run ping -t and see what happens.

 

[the54thvoid]

It's running - how long do i leave it?

Edit: ran for 9 mins - no drops, just one ping spike to 14ms.

I need to go work now.... meh....

 

[ixi]

On cmd use "ping xx.xx.xx.xx -t" and leave the ping for an about hour. Do two pings at the same time would be even better. One ping to your modem or router (dunno, which you have). and other to techpowerup for example. And check after 1 hour. If all pings will be sent and received then there is no fault in connection. While pinging try to avoid downloading/uploading stuff.

If you want to ping something in middle between your gateway and tpu. You can do as well "tracert xx.xx.xx.xx (techpowerup ip) -d" by thisbyou'll see IP without DNS names.

But as well, if you see PL between middle and not on end host it will mean that those routers in middle have low priority for icmp and it does not mean that there is packet loss.

On cmd use "ping xx.xx.xx.xx -t" and leave the ping for an about hour. Do two pings at the same time would be even better. One ping to your modem or router (dunno, which you have). and other to techpowerup for example. And check after 1 hour. If all pings will be sent and received then there is no fault in connection. While pinging try to avoid downloading/uploading stuff.

If you want to ping something in middle between your gateway and tpu. You can do as well "tracert xx.xx.xx.xx (techpowerup ip) -d"

But as well, if you see pl between middle and not on end host it could as well mean that those routers in middle have low priority for icmp and it does not mean that there is packet.


If after one hour you have PL to the end host example TPU via icmp (ping). Then from traceroute check those IP's in google to find which is your ISP last router before your packet goes into another ISP. As I'm from EU. I use ripe.net to check on IP and AS numbers which ISP it belongs to. Ping that last visible IP from your ISP and again TPU and check if there will be drops on both or only TPU. If only on TPU and your ISP router will answer without drops then it means most likely there is no fault in your ISP network. Could be problem in another ISP way till the host.



If you see already PL to your modem/gateway... well, you can try different lan cable, if new cable does not help then I would swap that swoopy device to another for testing purpose.

 

[the54thvoid]

Thanks for the reply. I'm going to give up for now.

Despite being told last month the issues would be fixed (the reason the 'engineer' came out after they weren't fixed), I've just been on to the call centre 4500 miles away and they have said there are ongoing widepsread issues which will "definitely" be fixed on 8th May.

So, I guess come 9th May, I'll be kissing goodbye to my 500 fibre, and saying hello to 67Mb max speeds (leave Virgin and go back to Sky). I'll just have to get used to those 150Gb games taking over 5 hours to download.

First world problems, I guess.

 

[Solaris17]

if you see PL between middle and not on end host it will mean that those routers in middle have low priority for icmp and it does not mean that there is packet loss.

This. the way to read MTRs in a watered down fashion is this.

The gateway shouldnt really drop, but thats the nature of wifi. We can tackle that later.

The route though will show loss. the carrier gateways and backbone connections do not care about your packets you are using cycles that they want to spend on traffic so they deprioritize when the ASIC reaches a certain load and drops you.

You need to look for sequencing. This is example of a good route.

Host > (no loss) > ISP > edge (loss) > transit router (no loss) > transit router (loss)

A BAD MTR

Host > (no loss) > ISP > edge (loss) > transit router (loss) > transit router (no loss)

You see how the ISP edge and the transit router has loss? loss in a row like that is generally a route problem. In that case the issue is probably the edge device. because the hop AFTER also has the same amount (within a few %) of loss.

these are fast and loose rules. If its your ISP you can maybe reset your ONT or modem for a diff IP and hope to god its on a network segment that isnt affected by that route. Otherwise its upstream and there isnt much you can do.

The phone tech wont be able to change a BGP route, that will be the NOC, you can let them know and call. but even if they manage to pass it on, it wont be done by supper time.


with stats like these

you can absolutely bitch. Even if they lose you as a customer, something is wrong with there transport links and they really need to look into it. There is loss along the entire route. how TF do you even load email??


gateway? can you try wired to run these MTRs? This makes me mad just to look at im going to go open anotheirn one I havent seen links this bad since seabone or the links to brazil

 

[the54thvoid]

I ran two cmd pings; one to my router, one to TPU. After a few minutes of both, TPU had five time-outs, my modem had no issues.

It's clearly Virgin's issue, and I don't think they have the will to fix it.

 

[eidairaman1]

I ran two cmd pings; one to my router, one to TPU. After a few minutes of both, TPU had five time-outs, my modem had no issues.

It's clearly Virgin's issue, and I don't think they have the will to fix it.

I was going to suggest looking at the coax F connectors to ensure they are fully seated and crimped on the coax wire and the connectors are tight to decices, I recall HDMI ports causing service issues on hard wired UVerse and solution was to move back to component cables. Heck a coax splitter can cause problems too

 

[Mussels]

Quite simply, you have a dozen steps along the way and you need to test all of them.
you need local ping tests between your PC and your router, your router and your first hop, and then onwards from there.

You can get local interference and packet loss, but you can also have problems with the PC itself (or another local networking device) - something like stutter from wallpaper engine isnt just visual, it can cause the whole PC to stutter and miss a packet or two


you look like you've got a solid connection from PC to router, so next is finding out if its the routers WAN connection, the modem (could be the same thing, with some devices), software settings (QoS settings can be set too aggressive, users could be smashing uploads or downloads and choking the connection, torrents could be running using thousands of connections a second, etc)

So i suppose disable wifi and run the one PC wired, and see if that changes anything as an isolation test - and if that doesnt help, then you need to try different cabling from the router/modem to the socket, and a different power source for it (even a different wall socket) to rule out power or ground issues, even if they are unlikely. Interally, LAN/WAN/WIFI are all different hardware pieces so some can be more sensitive to heat and electrical issues than others.


The "co-ax runs" sounds... odd? what's running over that?

 

[tabascosauz]

The "co-ax runs" sounds... odd? what's running over that?

DOCSIS providers run internet over coax/TV lines, via a feeder coax coming into the house. Kinda convenient cause you can connect whichever room(s) to the feeder and plug the modem in there, but well...... it's DOCSIS and sensitive both to the mandatory crappy modem's whims and the splitter config in the closet

 

[the54thvoid]

An update on some stuff. On advice from a friend, I tried the Virgin Media hub3.0 in modem mode.

This is the result of the MTR. It's not perfect - but it's a lot better.

The ldn-bb2-link.ip.twelve99.net has high packet loss on both modem and router - the rest are better, none above 3% loss.

And again - back on router, ten minutes later:

 

[tabascosauz]

An update on some stuff. On advice from a friend, I tried the Virgin Media hub3.0 in modem mode.

This is the result of the MTR. It's not perfect - but it's a lot better.

The ldn-bb2-link.ip.twelve99.net has high packet loss on both modem and router - the rest are better, none above 3% loss.

And again - back on router, ten minutes later:

You bridging to your own router now?

Is there any way to harass/escalate Virgin till you get a decent tech to come and look at it? I am once again stuck with a stupid DOCSIS provider again (such is the biyearly switcheroo forced by the internet duopoly here) but I am pretty fortunate that all these years their techs are actually really knowledgeable and always willing to talk about the tech. Problem with this fake fiber stuff is that so much of the hardware is not very visible or accessible to the user.

Also, isn't hub 3.0 like 2 generations out of date? I'd raise hell over that with Virgin. I think I was using hub3 like, 7 years ago lol

 

[Dragam1337]

With coax the issue is nearly always noise on the various QAM's and thus errors on downstream channels. Usually you can see how many correctable and uncorrectable errors you have by accesing the modem on 192.168.0.1 in a browser.

Correctable errors are fine with docsis 3.1, uncorrectable errors are not.

If you do have alot of uncorrectable errors, the issue will usually be the cabling from the fiber/coax switch box to your house.

 

[the54thvoid]

Yeah - I'm slowly realising from a lot of web searches that Virgin oversubscribe and sell the 'high speed fibre' bullshit without letting on their connections are flaky.

 

[Mussels]

Yeah - I'm slowly realsing from a lot of web searches that Virgin oversubscribe and sell the 'high speed fibre' bullshit without letting on their connections are flaky.

can you ditch whatever this coax stuff is, get a bridged modem setup at the leadin point and ethernet from there?

 

[eidairaman1]

With coax the issue is nearly always noise on the various QAM's and thus errors on downstream channels. Usually you can see how many correctable and uncorrectable errors you have by accesing the modem on 192.168.0.1 in a browser.

Correctable errors are fine with docsis 3.1, uncorrectable errors are not.

If you do have alot of uncorrectable errors, the issue will usually be the cabling from the fiber/coax switch box to your house.

If you have coax lines on the splitter not being used a shunt cap should be used to remove it from becomming an antenna for EMI. I've used them on the splitters too when I only had a 4 way splitter and not a 2 way.

I worked for AT&T 9 years ago when Uverse was in effect, they used HPNA for the coax runs going to TVs Hardwired and a Balun when going from UTP drop in the nid to Coax as the home run, otherwise ethernet was used, then a WAP was used for wireless STBs. Iirc Uverse came about in 2005 or 2006 and the last I saw Uverse used was in 2019.