@Si0mpe

Via the 10 Gig port, the sawtooth could come from queuing and/or flow control.

With the GF you could at best blow out the GF socket with compressed air to remove dust and thus further reduce the damping.

@Si0mpe only with [compressed air for Electronics,](https://www.brack.ch/kontakt-chemie-druckgasspray-dustoff-hf-67-340-ml-1116171?utm_source=google&utm_medium=cpc&utm_campaign=%21cc-pssh%21l-d%21e-g %21t-pla%21k1-bh%21z-baumarkt_hobby_smart_ margin_medium&utm_term=&adgroup_id=112188509418&ad_type=&prod_id=1116171&campaign_id=11301958057&SEMsc out&gclid=CjwKCAjws8yUBhA1EiwAi_tpER2YQ1gN3bCX0Tp1vfFGXKzGbsgyRNG_1v0TOB3Vzcncs-3YjmSZRhoCySQQAvD_BwE) because normal compressed air contains oil and water.

You can get rid of some dirt with compressed air, but not all.

You don’t clean your windows with compressed air, but use a cleaning cloth.

But I would rather use the here and this is normally used to clean the fiberglass.

Here is the official cleaning cloth for fiberglass:

and thus counter-checked if the measured values ​​are still not correct:

Fiber microscope handheld device:

@Si0mpe and everyone else 😊

I think I’ll shed some light on something dark:

• Why is a sawtooth visible in the US at CNLab?
  The subscription shaper for each Anschluss is stored in the central network element (BNG). If you have an 8.5gbps Anschluss, but only 1gbps subscription, then the packets are shaped away to the CPE (ONT) and discarded by the CPE. The latter causes the packages to be resend, which will result in a net throughput of the subscription speed. So if you place an average over the top values ​​of the sawtooth, you should get to 1g.

• Why does the result depend on the Ethernet port whether I see a sawtooth or not?
  This is really easy to understand, because here you are playing with a bottleneck that is sometimes 10g, 2.5g or 1g thick (thin). If I transport 10gbps from the PC via the CPE to the BNG, then in order to discard the 9gbps and tell the CPE to resend the packets, the bottom line is that the 1gbps is achieved.
  With the 2.5g port that’s already 4 times less and with the 1g port… not 10 times less but… that’s right, nothing at all. This means that neither shapers nor policers have to resort to anything because there is no more than 1gbps. And that’s why you don’t see any sawtooth there.

• Why does the router speed test show so much less?
  With the above information, this is also easy to understand, because there is no bottleneck called a switch or Ethernet connection to an Ethernet card in a PC/MAC. Therefore, the CPE crumbles out what it can and puts the policer in a lot of trouble because he has to discard 7.5 of the gbps/s sent and therefore have to kick off resends. In this case, this results in results that are far too low.

The whole topic is not new and we already had M subscriptions on FTTH and also on copper with fast lines, which reached up to 160mbps in the US. Here the M and L subscriptions in the US were limited to 100mbps and had absolutely the same effect. With copper we simply opened the US to 200mbps and the problem was gone (because no policer had to be used anymore).

Now the whole speed theater is only measurable/visible in the speed tests, but not in daily use. But you shouldn’t rule out the possibility that this could be visible when transferring very large files. Today, however, the customer network is still the limiting factor. Basically, no one builds an expensive 10gbps infrastructure at home and then reduces the speed subscription to 1gbps. That’s why there are no complaints, because the devices can rarely process more than 1gbps in the US today.

Nevertheless, 10gbps is becoming more and more common in the home network and your reactions show that there are speed-sensitive people who notice this. That’s why we want to get rid of this behavior, but it’s not that trivial. The experts are working with the supplier and we hope that there will soon be a solution that will “smooth out” the policer and that you can also see the full speeds in the speed tests.

Best regards

@Roger G

Thank you for all these explanations 😀

these would probably deserve to be covered separately in Knowledge Box

---

@Si0mpe wrote:

Hello everyone

Admittedly complaining at a high level.

Since the Inet M has now been raised to a level that is appropriate for the additional price and I couldn’t really use 10Gbit, I adjusted my subscription.

What surprises me a bit is that of the previous 7-8 Gbit/s that the Anschluss had delivered, there are now only 500-600 Mbit left for upload (now only 4 values ​​measured via the Internet box), while down is stable at 970Mbit/s

It’s hard for me to imagine that the neighborhood would suddenly take full advantage of the node.

So is it normal that you now have to backtrack from the promised 1Gbit symmetrical, or should I contact the hotline because there is a problem?

---

I have exactly the same starting point. I also switched from 10Gbit to Internet M, since practically everything at home is connected via WiFi anyway and I can only achieve a maximum of 600-700 Mbit/s in downloads anyway (limited by WiFi).

I like it - but only in the speed test! - also noticed that down is relatively stable at >900 Mbit/s, but upload is around 500 Mbit/s.

This is probably still an adaptation problem (software).?

Since braking connections requires extra effort, from a purely operational cost perspective, slow subscriptions should actually not be cheaper, but even more expensive than fast ones -)-)-)

Now all we have to do is convince marketing to simply let all connections run “open pipe” and to take into account the different “willingness to donate” of customers by using another distinguishing feature…

I could imagine, for example, the issue of reliability of the connection, e.g. subscription with/without a mobile failover stick, or other distinctions in the level of service that would allow differentiation of around CHF 10/month in the subscription price.

P.S.: Since the vast majority of customer connections are either mostly “idle” or with a low load, the total load in the Swisscom backbone would probably not even increase noticeably if all connections were actually running “open pipe”.

---

@User4500 wrote:
I have exactly the same starting point. I also switched from 10Gbit to Internet M, since practically everything at home is connected via WiFi anyway and I can only achieve a maximum of 600-700 Mbit/s in downloads anyway (limited by WiFi).

I like it - but only in the speed test! - also noticed that down is relatively stable at >900 Mbit/s, but upload is around 500 Mbit/s.

---

Please read a few posts about the information.

---

@WalterB wrote:

Presumably this is still an adaptation problem (software).?

---

@WalterB

See post #19

@Si0mpe

@Bergler

and simply to everyone with higher access speeds than the subscription: Please reboot the IB when you get the chance so that there is a new session and thus a new profile. Can you run the speed tests and compare them beforehand? I don’t think the sawtooth is gone, but real-world results are more helpful than internal ones.

Thank you and many greetings

---

@Si0mpe wrote:

was one of the first things I did, I didn’t have cnlab from before the reboot, because I only ever checked with the IB speed test - since it showed far too little, it was obvious to restart

The network is now connected to the 1GbE port, but it then “only” delivers 949Mbit/s instead of 998Mbit/s (in Ookla), but also in cnlab without any spikes

---

When did you do the reboot? We have only had the config active since last night, so I ask you to reboot today so that it works. Without any “jags” in the upstream, it sounds good.

Thank you, these are interesting tips. I’ll pass it on to IT. With the DS on the >1gbps ports you can see the new configuration (DS around 1.1 gbps thanks to the new shaper)