Mr Scratch
Expert Member
- Joined
- May 15, 2013
- Messages
- 4,872
- Reaction score
- 1,040
Could be local packet loss, which the VPN eliminates.
-
Join the MyBroadband community
South Africa’s biggest forum. Discuss, discover, and connect with thousands of members.
-
Question of the Day: What's the most interesting small town you've visited in South Africa?
Web Squad ISP
- Thread starter Gazed416
- Start date
- Status
It's definitly not a 'me' issue, as the issue would persist when running over a VPN, since the traffic still originates from my line, my switch, my computer. I have no packet loss without a VPN, nor with a VPN so it's certainly not that.
Web Client (Google Chrome) is the same, can't push anything higher than 35Mbit. Connect to ExpressVPN and I'm pushing full line speed.
There are also 3 or 4 Teraco datacenters in JHB now I believe, and I'm not sure if websquad peers with every one.
I suspect it has something to do with traffic routing within the websquad network, as that is the only thing I could think of that makes sense - since ExpressVPN would ignore websquad's routes in favor of it's own.
Mr Scratch
Expert Member
- Joined
- May 15, 2013
- Messages
- 4,872
- Reaction score
- 1,040
I can max my line from the web client using Web Squad, so it's not Web Squad (JHB, at least) so the issue could be you. Log a ticket and see what support says.
Mr Scratch
Expert Member
- Joined
- May 15, 2013
- Messages
- 4,872
- Reaction score
- 1,040
Also, routing could also be DNS. So making sure you're on 1.1.1.1 and that your system isn't reusing old cached DNS from previous servers is vital.
Mr Scratch
Expert Member
- Joined
- May 15, 2013
- Messages
- 4,872
- Reaction score
- 1,040
The VPN eliminates packet loss internationally, so no, it wouldn't persist. That's why CISP setup a local proxy so users with packet loss to band-aid the loss as the loss to local end-points is less punishing for TCP retransmits due to lower latency. Packet loss to EU is significantly more impact than packet loss locally. Test with some iperfs to really see if you have packet loss.
Mr Scratch
Expert Member
- Joined
- May 15, 2013
- Messages
- 4,872
- Reaction score
- 1,040
My restream test:
Are there any iperf servers in South Africa? I know seacom killed theirs last year...
Mr Scratch
Expert Member
- Joined
- May 15, 2013
- Messages
- 4,872
- Reaction score
- 1,040
Mr Scratch
Expert Member
- Joined
- May 15, 2013
- Messages
- 4,872
- Reaction score
- 1,040
Not necessarily internationally only, since a VPN would have their own routes corresponding to DC switch configurations/etc. The ISP routes would be wholly replaced once traffic has been routed from the ISP to the VPN and then the VPN routes it onward.
Cool, will test now.
Mr Scratch
Expert Member
- Joined
- May 15, 2013
- Messages
- 4,872
- Reaction score
- 1,040
Locally too but the difference will be less pronounced. And you're 100% right about the routing part. Web Squad is at Teraco so they get full access to pretty much every major player, but maybe some traceroutes will help with seeing the difference in routing.
This is my route to eu-london.restream.io with no vpn
The packet loss on 192.168.1.1 is the USG filtering an ICMP packet 'flood', so ignore that.
Edit: just re-ran the tracert and got a totally different route
Code:
|------------------------------------------------------------------------------------------|
| WinMTR statistics |
| Host - % | Sent | Recv | Best | Avrg | Wrst | Last |
|------------------------------------------------|------|------|------|------|------|------|
| 192.168.1.1 - 8 | 41 | 38 | 0 | 0 | 4 | 0 |
| as-vuma.jb-is-pld-01.za.ws.net.za - 77 | 13 | 3 | 0 | 6 | 11 | 2 |
| core.as-xe-02.jb1.za.ws.net.za - 0 | 54 | 54 | 0 | 1 | 38 | 1 |
| core.cr-xe-01.jb1.za.ws.net.za - 0 | 54 | 54 | 1 | 1 | 36 | 1 |
| core.pe-xe-ip01.jb1.za.ws.net.za - 0 | 54 | 54 | 1 | 1 | 37 | 1 |
| v101.core1.jnb1.he.net - 0 | 54 | 54 | 1 | 2 | 37 | 1 |
| 10ge3-11.core1.lon2.he.net - 0 | 54 | 54 | 157 | 159 | 211 | 159 |
| linx-224.as13213.net - 0 | 54 | 54 | 158 | 159 | 211 | 158 |
| 88.202.187.182.static.midphase.com - 0 | 54 | 54 | 157 | 184 | 355 | 257 |
| 109.123.74.58 - 0 | 54 | 54 | 157 | 160 | 211 | 158 |
|________________________________________________|______|______|______|______|______|______|
WinMTR v0.92 GPL V2 by Appnor MSP - Fully Managed Hosting & Cloud ProviderThe packet loss on 192.168.1.1 is the USG filtering an ICMP packet 'flood', so ignore that.
Edit: just re-ran the tracert and got a totally different route
Code:
|------------------------------------------------------------------------------------------|
| WinMTR statistics |
| Host - % | Sent | Recv | Best | Avrg | Wrst | Last |
|------------------------------------------------|------|------|------|------|------|------|
| 192.168.1.1 - 7 | 47 | 44 | 0 | 0 | 0 | 0 |
| as-vuma.jb-is-pld-01.za.ws.net.za - 92 | 12 | 1 | 0 | 6 | 6 | 6 |
| core.as-xe-02.jb1.za.ws.net.za - 0 | 56 | 56 | 0 | 2 | 18 | 5 |
| core.cr-xe-01.jb1.za.ws.net.za - 0 | 56 | 56 | 0 | 2 | 18 | 5 |
| core.pe-xe-ip01.jb1.za.ws.net.za - 0 | 56 | 56 | 1 | 2 | 16 | 3 |
| v101.core1.jnb1.he.net - 0 | 56 | 56 | 1 | 2 | 19 | 2 |
| 10ge3-11.core1.lon2.he.net - 0 | 56 | 56 | 157 | 163 | 197 | 158 |
| bbr01.lon01.networklayer.com - 0 | 56 | 56 | 157 | 159 | 192 | 159 |
| ae5.cbs01.tg01.lon01.networklayer.com - 0 | 56 | 56 | 158 | 159 | 170 | 159 |
| c1.13.2da9.ip4.static.sl-reverse.com - 0 | 56 | 56 | 158 | 159 | 187 | 159 |
| 7f.76.32a9.ip4.static.sl-reverse.com - 0 | 56 | 56 | 158 | 160 | 205 | 160 |
| cb.76.32a9.ip4.static.sl-reverse.com - 0 | 56 | 56 | 158 | 159 | 169 | 159 |
| eu-london6.restream.io - 0 | 56 | 56 | 158 | 158 | 168 | 160 |
|________________________________________________|______|______|______|______|______|______|
WinMTR v0.92 GPL V2 by Appnor MSP - Fully Managed Hosting & Cloud Provider
Last edited:
Mr Scratch
Expert Member
- Joined
- May 15, 2013
- Messages
- 4,872
- Reaction score
- 1,040
Try port 17001
Are you just proxying Coolidea's iperf lol?
Anyway the results are much of the same for iperf. Both results are equally as poor, connected over ExpressVPN or not. So I'm coming back to my routing theory lol.
Can't post results, because they are too long lmao, so ghostbin it is!
No VPN
Connected to VPN
Code:
PS E:\> tracert iperf-za.fbi.sh
Tracing route to trcvmh01.cisp.co.za [154.0.15.181]
over a maximum of 30 hops:
1 <1 ms * <1 ms 192.168.1.1
2 3 ms * * as-vuma.jb-is-pld-01.za.ws.net.za [160.119.228.1]
3 1 ms <1 ms <1 ms core.as-xe-02.jb1.za.ws.net.za [160.119.224.89]
4 4 ms 3 ms 1 ms core.cr-xe-01.jb1.za.ws.net.za [160.119.224.29]
5 1 ms 1 ms 2 ms core.pe-xe-ip01.jb1.za.ws.net.za [160.119.224.19]
6 4 ms 3 ms 3 ms coolideas.ixp.joburg [196.60.8.171]
7 4 ms 2 ms 1 ms uzv-cust.coolideas.co.za [154.0.5.11]
8 21 ms 21 ms 21 ms c2m-backbone.coolideas.co.za [154.0.2.94]
9 152 ms 21 ms 21 ms 154.0.2.58
10 21 ms 21 ms 21 ms 154.0.15.181
Trace complete.Anyway the results are much of the same for iperf. Both results are equally as poor, connected over ExpressVPN or not. So I'm coming back to my routing theory lol.
Can't post results, because they are too long lmao, so ghostbin it is!
No VPN
Code:
https://ghostbin.com/paste/o7gdt
Code:
https://ghostbin.com/paste/whow3
Last edited:
Mr Scratch
Expert Member
- Joined
- May 15, 2013
- Messages
- 4,872
- Reaction score
- 1,040
Since you couldn't connect to mine, yeah. Gotta troubleshoot your stuff lol
Run with
Code:
iperf -R -u -b 90M -p 17001 -c iperf-za.fbi.sh
Code:
iperf -R -u -b 90M -p 17001 -c iperf-uk.fbi.sh
Code:
https://ghostbin.com/paste/c4pg4
Code:
https://ghostbin.com/paste/4re76Tests done connected directly to the CPE. RIP packet loss, though I'm not sure how stable CISP's iperf is hahaha, so I'll put it down to that, given how the UK iperf is stable...
Though packet loss wouldn't explain why ExpressVPN routes with no issue in any event, since the traffic still egresses over my vuma line.
Mr Scratch
Expert Member
- Joined
- May 15, 2013
- Messages
- 4,872
- Reaction score
- 1,040
It's stable enough to detect packet loss well, and as suspected, your have loss. The explanation as to why a VPN eliminates the TCP re-transmit penalty is because packet loss penalty is in proportion to the latency between you and your endpoint. Your latency to the VPN is lower than your latency to the Restream server, so the VPN acts as a bandaid for the loss as the VPN has no loss to Restream. Think of it as passing traffic to the VPN, locally, and the VPN does everything else beyond that so the loss you have on the local loop doesn't affect traffic internationally.
https://www.netcraftsmen.com/wp-content/uploads/2014/12/20120410_Impact-of-packet-loss.pdf
Test this by connecting to a VPN server in UK, and retesting with Restream, and let me know the results.
Since this is directly over the CPE, how the hell is it fixed? Networking was one of those units I didn't particularly enjoy at University... The thing I don't get is that packet loss would affect my download as well, but that is fine.
Also tried different cables from PC to CPE (cat 5e, cat 6 and cat 7) and different ports, same results. Also tried on my desktop and laptop for good measure, same results.
Mr Scratch
Expert Member
- Joined
- May 15, 2013
- Messages
- 4,872
- Reaction score
- 1,040
Log a fault with Web Squad, give them all the information and tests you've done and tell them they need to log a fault with your line provider.
- Status