Diagnosing performance/throughput issues

RouterOS General
1

Hi Forum,

this is my first post in this forum and I am well aware that there might not be a straight solve-all answer to my questions.

First I would like to explain my setup.
I am running a WISP with ~20 clients and CCR1009 at the core.
I using PPPOE for client connectivity and user-manager on the core router.
My outbound bandwidth consists of the following lines:

  • 2x dedicated (one customer only) 16Mbit/1Mbit DSL lines
  • 1x shared (all other customers) 100Mbit/12Mbit Cable (DOCSIS3.0) line

Unfortunately the cable line employs (enforces) the use of an additional router with NAT (FritzBox Cable). To avoid having to many NAT stages I disabled NAT in the core router a using static routes inside the FritzBox.
This all is working fine in principle.

I should mention, that I also have a Linux box where the core router is and where all the uplinks are located.

Now I observe the following:

  1. I can download a file (over WLAN) from my Linux box with the full speed the WLAN link provides (3MByte/s for example).
  2. I can download a file from the internet to my Linux box (over 1Gbps Ethernet link) with full speed of the cable uplink (10 MBytes/s)
  3. When I try to download a file from the internet over WLAN, I get fluctuating transfer rates of about 1MByte/s.

I experimented with the following:

  • Disabling various firewall rules (no change in throughput)
  • Disabling queues completely (no change in throughput)
  • Creating a OVPN tunnel (also over WLAN), that is bridges into the local subnet of the FritzBox directly (3MBytes/s as desired)

So my questions are:

  • Can you give me any pointers to look for/ to read up on/ to try out?
  • Is there any problem with pppoe links and bandwidth beyond 1MByte/s?
  • Can I “prove” that the bottleneck/problem is not caused by the core router?

Thanks for all hints in advance.

2

Hi,

if you only experience this issue over PPPoE, then MTU could be the cause.
Whats your path MTU? What MTU is configured for the PPPoE connection?

Ape

3

Hi,

I checked my MTU the following way (over wlan, PPPOE and cable connection):

ping google.com -f -l 1384

Pinging google.com [173.194.113.4] with 1384 bytes o
Reply from 173.194.113.4: bytes=1384 time=26ms TTL=5
Reply from 173.194.113.4: bytes=1384 time=28ms TTL=5
Reply from 173.194.113.4: bytes=1384 time=27ms TTL=5
Reply from 173.194.113.4: bytes=1384 time=31ms TTL=5

Ping statistics for 173.194.113.4:
Packets: Sent = 4, Received = 4, Lost = 0 (0% lo
Approximate round trip times in milli-seconds:
Minimum = 26ms, Maximum = 31ms, Average = 28ms


ping google.com -f -l 1386

Pinging google.com [173.194.113.4] with 1386 bytes of data:
Reply from 192.168.100.1: Packet needs to be fragmented but DF set.
Packet needs to be fragmented but DF set.
Packet needs to be fragmented but DF set.
Packet needs to be fragmented but DF set.

Ping statistics for 173.194.113.4:
Packets: Sent = 4, Received = 1, Lost = 3 (75% loss),

So my link MTU should be 1384 + 28 = 1412, and this is what I set my PPPOE server config to for max MTU und max MRU.


I already tried (just for fun) some unreasonable value (like 1000) and this did not affect the issue unfortunately.

4

Packet capture at different points and see if the traces tell you anything interesting. I’d start with what a client sees when talking to the Internet.