International connections being throttled?

A couple of months ago I noticed that download speeds to servers outside of Australia had slowed down immensely.
I didn't think much of it as the Optus outage map showed that work was being done on the towers in my area.

The work is finally over but I am still having problems with downloading.
Connections outside of Australia hover between 50-70KB/s. I can make multiple connections to speed up the transfers, but I didn't need to do this in the past.

The Vividwireless thread on Whirlpool reports multiple users with congestion issues, but nothing that matches this behaviour.

Connections to Australian servers are fine, and were running at decent speeds even while work was being done.

Any input on whether this is due to congestion that people reported, or is Optus doing something dodgy upstream?

- consty

Comments

#1

ozzcaddy-MODERATOR wrote 5 days ago

consty,

Your comment about 'Optus doing something dodgy upstream', Vividwireless is a separate entity to Optus even though Optus owns Vividwireless. They use separate antenna's on the same tower and are separate networks, .

Can you provide the overseas websites that your having problems with, so it can be looked at.

What Suburb/Postcode are you in?

#2

consty wrote 5 days ago

Hey Oz,

4chan.org is one of the sites that used to load fine for me, and I was resorting to browsing on my phones connection.
Rsync transfers from a VPS in France started having issues and timeouts.
The archive.ubuntu.com repo was slow so I switched to an aussie one.
Youtube and Twitch were hit and miss. Some videos load fine while others would buffer constantly.

I'm in Applecross, 6153. Three of the towers had work done, one of which had the resolution ETA extended multiple times.

- consty

#3

ozzcaddy-MODERATOR wrote 4 days ago

Thanks consty for info asked.

I'm in inner Sydney and will do some traces from here to a couple of the websites you provided. I have been on YouTube recently pretty frequently, and have not had any problems, and that is with slow speeds at the moment of 2.5Mbps but with low ping.

I just looked at the Optus Network Status map, and if you enlarge the map you will see you are totally surrounded by Orange tower icon which indicate work is being done or is to be done on those towers and there is one tower (Red) with a problem to the East. You can click on the tower icon to find out more info of dates and times.

I think with all these towers having work done, customers have been pushed on to other towers including your towers which would cause congestion. Optus maintains the towers no matter if it relates to Vividwireless or Optus.

I'll post back results of the traces later.

#4

ozzcaddy-MODERATOR wrote 4 days ago

Below are results of 2 traces to websites 4chan.org and archive.ubuntu.com as provided by consty.

As both traces ran, they basically stopped at Line 3 and was slow thereafter.

Frosty has more knowledge of interpreting the results of the traces. If necessary, he has contacts at Vividwireless if needed to be looked into.

Microsoft Windows [Version 10.0.17134.112]
(c) 2018 Microsoft Corporation. All rights reserved.

Tracing route to 4chan.org [104.16.120.221]
over a maximum of 30 hops:

1 1 ms 1 ms 1 ms homerouter.cpe [192.168.0.1]
2 68 ms 152 ms 59 ms sydequbr06.bbn.unwired.net.au [220.101.190.2]
3 59 ms 85 ms 39 ms 220.101.190.206
4 1895 ms 60 ms 754 ms ge-2-3.br2.equ.syd.bbn.unwired.net.au [220.101.190.198]
5 1514 ms 108 ms 68 ms as13335.nsw.ix.asn.au [218.100.52.11]
6 184 ms 62 ms 57 ms 104.16.120.221

Trace complete.

Tracing route to archive.ubuntu.com [91.189.88.149]
over a maximum of 30 hops:

1 2 ms 1 ms 1 ms homerouter.cpe [192.168.0.1]
2 53 ms 43 ms 41 ms sydequbr06.bbn.unwired.net.au [220.101.190.2]
3 43 ms 37 ms 30 ms 220.101.190.206
4 66 ms 25 ms 29 ms 220.101.73.153
5 * * * Request timed out.
6 59 ms 27 ms 48 ms 59.154.18.12
7 324 ms 405 ms 406 ms 203.208.191.21
8 377 ms 411 ms 401 ms 203.208.172.174
9 382 ms 406 ms 407 ms 203.208.171.114
10 390 ms 412 ms 401 ms ae-1-3114.edge5.London1.Level3.net [4.69.148.218]
11 413 ms 407 ms 406 ms SOURCE-MANA.edge5.London1.Level3.net [212.187.138.82]
12 418 ms 407 ms 413 ms danava.canonical.com [91.189.88.149]

Trace complete.

#5

Frosty-MODERATOR wrote 4 days ago

I am at a disadvantage from a testing perspective, because I am in the UK at the moment and won't be back until August. So I can't really do any comparative testing. Having said that …

I'm on an ADSL link in the north of the UK and ran a tracert to 4chan.org with the following result:

C:\>tracert 4chan.org
Tracing route to 4chan.org [104.16.120.221]
over a maximum of 30 hops:
1 2 ms 1 ms 8 ms dsldevice.lan [192.168.1.254]
2 71 ms 19 ms 20 ms 410.xe-2-2-0.central10.psb-bng04.plus.net[84.93.253.120]
3 24 ms 19 ms 21 ms 411.be8.psb-ir02.plus.net [84.93.253.127]
4 32 ms 49 ms 22 ms 195.99.125.138
5 22 ms 34 ms 44 ms peer8-et-0-0-2.telehouse.ukcore.bt.net [62.172.103.172]
6 20 ms 19 ms 22 ms 109.159.253.253
7 31 ms 19 ms 20 ms 104.16.120.221
Trace complete.

That tells me that 4chan.org is "distributed", in that from a UK perspective the servers for 4chan.org are in the UK. Similarly, for Australia we go to a Sydney-based datacenter.

I would suggest for starters setting up 2 concurrent pings:

1. ping 4chan.org -t
2. ping sydequbr06.bbn.unwired.net.au -t (or whatever is the 2nd hop upstream of you, the first point at which you hit the vividwireless/unwired network)

Leave both running. If both give you packet loss or poor performance at the same time, then it is either your modem-to-vividwireless or just the tower at vividwireless which is the problem. If the 4chan.org shows bad pings, but the modem-to-vividwireless looks okay at the same time, then you can conclude that the problem is upstream of the tower somewhere.

This sort of test is not conclusive, but it can help point in the right direction at least. Give that a try and see what you can learn. Then we'll know whether to focus on your modem/tower or on the vividwireless core network or their links to the outside world.

#6

consty wrote 4 days ago

Latency to the servers in question are no different from when things were running fine.

4chan goes via Cloudflare, which responds quickly enough.

Tracing route to 4chan.org [104.16.119.221]
over a maximum of 30 hops:

1 <1 ms <1 ms <1 ms 172.16.10.1
2 1 ms <1 ms <1 ms 192.168.10.1
3 15 ms 23 ms 24 ms 180.216.250.1
4 36 ms 25 ms 22 ms 10.112.0.245
5 42 ms 84 ms 79 ms 198.32.212.107
6 33 ms 39 ms 30 ms 104.16.119.221

But loading files via them are incredibly slow, but weren't before (again, between 50-70KB/s).

Running some traceroutes to the other servers mentioned does yield something interesting.

Tracing route to archive.ubuntu.com [91.189.88.152]
over a maximum of 30 hops:

1 2 ms 2 ms 2 ms 172.16.10.1
2 1 ms 1 ms <1 ms 192.168.10.1
3 26 ms 26 ms 26 ms 180.216.250.1
4 30 ms 28 ms 34 ms 10.112.0.245
5 19 ms 19 ms 31 ms 119.225.48.173
6 * * * Request timed out.
7 * * * Request timed out.
8 * * * Request timed out.
9 87 ms 89 ms 79 ms 59.154.18.22
10 250 ms 250 ms 229 ms 203.208.190.69
11 * * 239 ms lag-8.ear2.SanJose1.Level3.net [4.68.73.69]
12 379 ms 381 ms 376 ms ae-1-3114.edge5.London1.Level3.net [4.69.148.218]
13 371 ms 442 ms 402 ms SOURCE-MANA.edge5.London1.Level3.net [212.187.138.82]
14 373 ms 377 ms 367 ms steelix.canonical.com [91.189.88.152]

Tracing route to (VPS in France)
over a maximum of 30 hops:

1 <1 ms <1 ms <1 ms 172.16.10.1
2 <1 ms <1 ms <1 ms 192.168.10.1
3 44 ms 25 ms 30 ms 180.216.250.1
4 31 ms 28 ms 78 ms 10.112.0.245
5 22 ms 38 ms 26 ms 10.112.0.254
6 33 ms 17 ms 26 ms 119.225.139.57
7 * * * Request timed out.
8 * * * Request timed out.
9 * * * Request timed out.
10 * * * Request timed out.
11 93 ms 76 ms 78 ms 59.154.18.6
12 248 ms 238 ms 238 ms 203.208.191.169
13 240 ms 226 ms 243 ms 203.208.172.174
14 225 ms 249 ms 229 ms 203.208.171.114
15 * 390 ms 381 ms ae-2-3203.ear3.Paris1.Level3.net [4.69.161.126]
16 388 ms 382 ms 414 ms 212.3.235.198
17 380 ms 401 ms 394 ms 195.154.1.189
18 * * * Request timed out.
19 * * * Request timed out.
20 * * * Request timed out.
21 * * * Request timed out.
22 * * * Request timed out.
23 385 ms 385 ms 404 ms (VPS in France)

Whenever the trace hits anything in the 59.154.18.x subnet (*cough* Optus *cough*) the latency spikes.
The next hop after is via Singapore, but the ping times were like that before I experienced the slow-downs.
I used to have ping times of around 400ms to my VPS, but would almost max out my connection on a single-threaded transfer.

Again, I'm only getting throttled transfers for servers outside of Australia.
I could be getting slow speeds within Australia, but I haven't downloaded anything from a service that may or may not be peering with Optus/Vivid/whomever.

- consty

#7

Frosty-MODERATOR wrote 3 days ago

Thanks for the detailed info. Okay, if we're not able to identify the issue by means of packet loss using ping/tracert, then we're in the realms of prioritisation I guess.

It has been a while since this was discussed here, so it is always possible that things have changed a bit. But I will tell you what I know (or think I know).

******************************************************************
EDITED: just removed some out of date info. See below.
******************************************************************

I've checked with the network engineering team. There's no Quality of Service (QoS) being applied these days to various kinds of traffic. The only "thing" to be aware of is that Peer to Peer (P2P) traffic such as torrents and similar is limited, which is fair enough.

Is it possible that your traffic is of a type that might be identified in the P2P category? Really large file transfers and so on?

In the past, using a VPN to tunnel traffic point-to-point was an effective way of dealing with individual cases where traffic management was interfering with a particular user's traffic, so you might want to consider that for your VPS and see if it makes a difference.

#8

consty wrote 3 days ago

Hey Frosty,

The bulk of the traffic is http/https, and rsync transfers over ssh (port 22, TCP, to and from one IP).

I had no problems doing this for years. Only recently has it been an issue.

There is a VPN tunnel between my machine and the VPS, but it was always faster (and still is) to transfer directly (TCP vs TCP over TCP/UDP).

- consty

#9

Frosty-MODERATOR wrote 2 days ago

The only other thing I can think of then, apart from the traffic being wrongly identified as P2P, would be if the limitations are being applied outside of the vividwireless network, somewhere upstream.

Years ago there were problems with connections not getting through at all for some 180.216.0.0/16 addresses because in years gone by they'd been associated with spam, and this was happening internationally. So there is some precedent, but that wasn't a rate-limiting thing, it was an all-or-nothing you either can or cannot connect thing.

I wonder if Optus are limiting vividwireless bandwidth in some way? Either a per-connection limit, or just in general? It makes sense that there would be a limit, but I'm afraid that I don't have any means of testing that idea.

#10

Frosty-MODERATOR wrote 8 hours ago

I have had a chat to the network engineering team. As far as the interconnects are concerned, getting from the vividwireless core network out to the rest of the world, they are very confident that there is not a bandwidth issue there. They did share with me some of the details, though I cannot share that info in this forum. It was pretty convincing.

So it may be that there is a problem upstream somewhere.

If you can find a choke point, where the ping time suddenly leaps upward in an unusual manner, then this can be followed up.

Warning
In participating in this forum you agree not to post or authorise others to post any material that is in our opinion offensive, defamatory, inaccurate, harassing, threatening, invasive of the privacy of others, sexually explicit or illegal. You also agree that you will not post or authorise others to post any copyright material that is not owned or licensed to you. In the event of any breach of this Policy we reserve the right to remove the post without notice to you.

Disabling comments and removing posts
Should we consider a thread or comment in violation of this Policy, we reserve the right to immediately remove part of or entire posts. We also reserve the right to disable, remove or archive posts that have been dormant for 14 days.

Member Login