Categories
data

Hunting Network Speed

I’ve had a consumer UPC internet connection for years now. The product should deliver 75Mbit/s download and 7.5 Mbit/s upload speed.

I am monitoring the performance of my bandwidth automatically. So when UPC forced a modem ‘upgrade’ on me, I immediately knew that the performance of my connection dwindled. That was at the end of January 2018. A, until then, somewhat decent bandwidth of an average of 71 Mbit/s came down to about 40 Mbit/s.

The old cable modem was a modem only device. It was black and wall-mountable. The new modem is a white designer piece and offers full router and WiFi capabliliy. It is not wall mountable and that is slightly impractical. But I am using this WiFi capable cable modem as modem only. There is a setting that allows simple modem use, which disables all WiFi and smart-router features. I’ve my own router behind it.

The performance issues are clearly visible, after the modem has been switched at the end of January 2018

The graph clearly shows the drop in performance after the modem ‘upgrade’. One can only speculate what the reason for a little recovery in July is. In July of 2018 the performance came back – which could be due to less usage in the network, as it coincides with school holidays. I do have no proof for that though.

The speed drop is nicely seen in a heat map that plots speed by hour of the day over months.

As the speed did not improve, even after hours spent with the hotline, and multiple resets and even an active cooling attempt – I opted for another modem switch. Same type of modem – but I switched everything: modem, power supply, cables etc. This was at the beginning of December 2018.

It had no effect.

Not even actively cooling the modem did help.

To make things worse, the modem worked fine for about 12-18 hours after a power cycle. So the hotline would tell me to power-cycle the modem and connect my laptop directly to it, to test for speed. Of course it worked fine then:

classic manual power cycle at the start at 12:00, after that a stable connection for 16 hours, then an instable connection. Another manual power cycle the next day at 19:00. The firmware upgrade happened at March 25th at 12:00.

I kept mentioning the issue to a friend, who has connections to T-Mobile (UPC). He mentioned that it might not be a hardware but a firmware problem. Why the support hotline didn’t know that I will never know. I opted for a firmware update of the modem – and voilà – the connection is where it should be:

Download and upload-speed from January 2018 to Beginning of April 2019
Heatmap of Downloadspeed from January 2018 to Beginning of April 2019

It seems UPC has been distributing modems with (ancient) firmware (April 2017), that don’t fit their infrastructure any more. And instead of pushing a new working firmware to new modems, they swap brand new modems and keep customer satisfaction artificially low.

I’m happy the issue is solved now. But I’d rather spend less time debugging issues that my provider should fix on it’s own. I can’t wait for legislation that would allow customers to run their own modems!

You can find my data analysis for this blog post here (2017-09 to 2018-08) and here (2018-01 to 2019-04).

Categories
english Howto

Hetzner Root Server Networking Configuration…

I’ve been setting up a new server at hetzner.de.
I ran into problems when configuring the network. The server is running Debian (wheezy).

hetzner network info
note the last line: “The additional route to the gateway is now no longer necessary.” not only that: it will not work.

The basic configuration looked like this:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
## /etc/network/interfaces example Hetzner root server
# Loopback-Adapter
auto lo
iface lo inet loopback
#
# LAN interface
auto eth0
iface eth0 inet static
# Main IP address of the server
address 192.168.0.250
# Netmask 255.255.255.255 (/32) independent from the
# real subnet size (e.g. /27)
netmask 255.255.255.255
# explicit host route to the gateway
gateway 192.168.0.1
pointopoint 192.168.0.1

This should work, as mentioned in the Hetzner DokuWiki.

I added DNS servers at the end (use your DNS servers here or pick an open DNS server)

1
dns-nameservers X.X.X.X Y.Y.Y.Y

at the end since I’ve resolvconf installed.
eth0 did not come up correctly.

When trying ‘ifdown eth0; ifup eth0’ I kept getting:

1
2
3
ifdown: interface eth0 not configured
RTNETLINK answers: File exists
Failed to bring up eth0.

This error would show up at boot time or when trying to start eth0 by hand.
The setup would look fine otherwise, IP was correct network seemed to work, but the DNS-servers were not added correctly. Weird!

‘ifdown –force eth0; ifup eth0’ worked. Server went off for a second but came back. with DNS-servers setup correctly. Interesting!

I started to comment out lines from /etc/network/interfaces.
Et voilá!

It turns out: It is deadly to try to configure ‘gateway’ in /etc/network/interfaces!

Finally I used this:

1
2
3
## /etc/network/interfaces working Hetzner root server
auto lo
iface lo inet loopback

auto eth0
iface eth0 inet static
address 192.168.0.250
netmask 255.255.255.255
# next line optional
network 192.168.0.0
## never EVER use the next line! you have been warned!
## gateway 192.168.0.1
pointopoint 192.168.0.1
dns-nameservers X.X.X.X Y.Y.Y.Y

I hope this post will save others some time to fix this issue with their setup.