Poor ML/PPP Upstream Performance

RouterOS General
1

I’m not sure how long this has been going on, but it’s driving me nuts.

Hardware: Mikrotik RB450G running RouterOS 4.4, Thomson ST516 v6 dsl modem
Line: I’m currently testing in my lab, which has a single dry-loop
Setup: I have the Mikrotik setup as a NAT gateway for this test with a MacBook behind it

Steps to reproduce…

Fast configuration:

[asherman@shermangw] /interface pppoe-client> print
Flags: X - disabled, R - running 
0  R name="pppoe-out1" max-mtu=1484 max-mru=1484 mrru=disabled interface=dsl1 user="********" password="********" 
     profile=default service-name="" ac-name="" add-default-route=yes dial-on-demand=no use-peer-dns=yes allow=pap

Note value of the “interface” parameter. I now do a few different speed tests and get 5.12 by 0.66 Mb/s.

I then make the following change:

[asherman@shermangw] /interface pppoe-client> set interface=dsl1,dsl2 pppoe-out1 
[asherman@shermangw] /interface pppoe-client> print
Flags: X - disabled, R - running 
0  R name="pppoe-out1" max-mtu=1484 max-mru=1484 mrru=disabled interface=dsl1,dsl2 user="********" 
     password="********" profile=default service-name="" ac-name="" add-default-route=yes dial-on-demand=no 
     use-peer-dns=yes allow=pap

Notice the new value of the “interface” parameter. (There is still only one modem plugged in though.) A couple more speed tests and I get 5.11 by 0.10 Mb/s. So the downstream is fine but the upstream just completely goes to crap.

I have tried this on multiple dry-loops using multiple Mikrotik units and even with multiple DSL accounts. There has to be a reason for this.

Since we really need the bandwidth for our customers, I hope someone has some ideas.

Thanks!

A.

2

It’s too bad that the performance is degraded when one link goes down… But how often will this be the case? In normal operation won’t all links be up? I’m guessing it does a sort of round-robin transmit and when one link fails it tries to use that interface anyway and has to resend on the other link.

3

Performance is also degraded with multiple links going, but I brought it down to a single link to isolate the issue.

A.

4

Well the performance isn’t categorically bad - we have mlpppoe solutions deployed at several customers, one of them even bonding three 6176/512 lines without a problem. I would rather use 3.30 for MLPPP however.

And as an additional Info: When one link of a MLP Bundle fails, the whole connection dies for about 30secs to 3 Minutes, afterwards it should continue to work normally on one link.

5

Could my problem be configuration?

6

Mine looks like this:

0 R name=“pppoe-out1” max-mtu=1492 max-mru=1492 mrru=1524
interface=ether2,ether3 user=“" password="”
profile=default service-name=“” ac-name=“” add-default-route=no
dial-on-demand=no use-peer-dns=no allow=pap,chap,mschap1,mschap2

The only real differences are MTU and MRRU…
You might try different values for MTU/MRU and play around with MRRU settings.
Default MRRU if your provider’s LAC is Cisco-based should be 1524.

7

Thanks. I thought the max parameters were just that, maximums, and that the system was automatically adjusting them. I’ll try adjusting them anyway.

A.

8

At least in my experience those max values get overridden sometimes.
eg. the cisco router we use as LAC pushes MTU 1524 when using MLPPP and the mikrotik accepts it, even though max-mtu 1492 is set.
if I then don’t set MRRU to 1524, an MTU asymmetry happens, which may cause problems with large packet sizes and maybe those interesting upload problems you experience. When I view the Status of the pppoe-out1 interface, it shows MTU and MRU both as 1524.