Thaffic Shaper issue

[muchacha_grande]

I have an issue with Traffic Shaper.
I migrated from pfSense to OPNSente with the exat same configuration.
But whet I needed to port thaffic shapping, I realized that there is no way to use dynamic IPs, as with aliases in pfSense.
So I can't make things work just as before migrating because I have a PBX behind the fw and some dynamic IP locations to prioritize.
And on the other side, the best way I found to achieve a good VoIP quality is using limiters because queues did not perform well even using a 1-100 priority between VoIP and data traffic.
This fact left me with a lost of speed because I have a 12 Mbps cablemodem connection with a turbo speed of 30 Mbps (on the first seconds) and limiters have to be set to, lets say, 11 Mbps to grant VoIP have a 1 Mbps channel.

EDIT:
After writting this, I searched throug the forum and realized that ipfw does not filter packets as pf does, so there is no way to achieve what I need.
Also, I found someone with the same problem as mine when using queues. Priorities are treated different than in pfSense, so VoIP doesn't perform well using queues leaving limiters as the best way of qos.
I shoulded search in forum before posting this issue. Sorry.

[franco]

Hi there,

Also, I found someone with the same problem as mine when using queues. Priorities are treated different than in pfSense, so VoIP doesn't perform well using queues leaving limiters as the best way of qos.

I'm not sure I understand. The technology for the "shaper" in OPNsense is what is used as the "limiter" in pfSense.

What kind of dynamic IPs are we talking about?


Cheers,
Franco

[muchacha_grande]

Hi Franco,

with dynamic IP I mean, IP addresses that may change, so using aliases and hostnames IPs are resolved every 5 minutes (I think) and if some IP change the alias takes the new IP and the limiter could be filtered using it automatically.

The limiters are right for many cases, but I really use queues with priorities in pfSense with an excellent result.
Priorities on OPNSense seems to work different and I noticed some glitches with VoIP using queues, even if I use priority of 100 for VoIP vs 1 for data.
  Using queues I can always use all of the bandwith, including the turbo on the first seconds.
  If I use limiters I just loose the turbo and my data is limited even if I'm not using the phone.

Cheers,
Francisco.

[muchacha_grande]

Hi Franco,

finally after all my testings on OPNSense traffic shaper, I didn't get the same performance compared to pfSense with the exact same configuration.
I think I'll need to get a deeper understanding on the involved parameters on OPNsesnse sinse may be a problem of configuration.
I tested, limiters and queues and none of them solved the voip choppy voice.
I there a chance of implementing on OPNSense the same traffic shaper mechanism that pfSense uses?
By now that's the only flaw I can find on OPNSense sinse everything else is really much better.
I implement VoIP PBXs at my customers and in almost every case I use QoS to get better results.
My plans are moving to OPNSense and eventually migrate my customers pfSense routers.

Cheers

[franco]

Hi there,

The reason we removed the original shaper was that the framework named ALTQ came from OpenBSD but was later removed there, FreeBSD still has it, but always off by default, scheduled for final removal in maybe FreeBSD 12 because of drawbacks in the implementation. It also had a few associated custom patches. That made it easier to make the decision to leave it out. It won't come back.

If the performance is not up to specs, you could share your numbers and test methodology to help find the underlying issue.


Cheers,
Franco

[muchacha_grande]

Hi,

  after I posted my previous message I read about the traffic shaper on FreeBSD.

  The only test I've made was based on the day by day use of a PBX I have behind OPNSense.

  Voice had glitches and it was difficult to have a normal conversation.

  Compared to my previous firewall, it was a disadvantage.

  Now, to make some meassurements I need some advices because I don't know what kind of tests would be useful to the developers.

  Could you give me some guidelines to follow and make the right tests?

Thank you in advance,
Francisco.