Goedendag,
Wil voor een huawei ont eigen apparatuur gebruiken, omdat deze kuren vertoont en regelmatig de verbinding wegvalt.
In de gedachte dat zowel een nieuwe ont (huawei) of eigen apparatuur opnieuw geregistreerd dient te worden na de huidige ont uit het systeem te hebben getrapt en de wetgeving toelaat het gebruik van eigen apparatuur, wil ik hier een lijntje uitgooien om gebruik van eigen apparatuur voor elkaar te boksen, middels onderstaande module.
+9
Hi
Hi
Ik heb met de XGS-ONU-25-20NI in mijn UDM-Pro veel last van bufferbloat helaas, met het ONT ook al wel last van maar minder dan nu. Tijdens downloaden moet ik zelfs wachten op het laden van een pagina op mijn lokale netwerk dit is niet goed. Ik mis mijn KPN AON lijn nu al..
Meer mensen met dit issue? Of mensen die ook een XGS-ONU-25-20NI hebben in hun UDM-Pro die dit probleem niet hebben?
Thanks!
EDIT: Door de SFP+ nu op 1Gbps FDX te zetten lijkt het probleem weg, de bufferbloat is nog niet perfect maar fatsoenlijk surfen lukt weer tijdens het downloaden!
I also have bufferbloat issue with Zaram module in SFP+ cage (cage on X710 used by virtual pfSense router)
Looking at Huawei ONT, this should be solvable/configurable. In my case, for time being, I just limited RX rate on x710 NIC, which took care for most of the bufferbloat. But not all.
That’s unfortunate to hear I was hoping this was an issue with the UDM-Pro, but it seems it’s an issue with SFP+ and Odido in general for some reason. I am still using the Huawei ONT as long as I’m not upgrading to a higher speed. Let me know if you find a permanent solution to this issue, and thank you for testing.
Hi
Thanks, ga ik doen!
Ik heb met de XGS-ONU-25-20NI in mijn UDM-Pro veel last van bufferbloat helaas, met het ONT ook al wel last van maar minder dan nu. Tijdens downloaden moet ik zelfs wachten op het laden van een pagina op mijn lokale netwerk dit is niet goed. Ik mis mijn KPN AON lijn nu al..
Meer mensen met dit issue? Of mensen die ook een XGS-ONU-25-20NI hebben in hun UDM-Pro die dit probleem niet hebben?
Thanks!
EDIT: Door de SFP+ nu op 1Gbps FDX te zetten lijkt het probleem weg, de bufferbloat is nog niet perfect maar fatsoenlijk surfen lukt weer tijdens het downloaden!
I also have bufferbloat issue with Zaram module in SFP+ cage (cage on X710 used by virtual pfSense router)
Looking at Huawei ONT, this should be solvable/configurable. In my case, for time being, I just limited RX rate on x710 NIC, which took care for most of the bufferbloat. But not all.
That’s unfortunate to hear I was hoping this was an issue with the UDM-Pro, but it seems it’s an issue with SFP+ and Odido in general for some reason. I am still using the Huawei ONT as long as I’m not upgrading to a higher speed. Let me know if you find a permanent solution to this issue, and thank you for testing.
I made longer writeup about this and testing here
But, long story short, Huawei is also suffering from it. While they certainly have SQM, which allows to reach maximal throughput, it still suffers heavily from crazy latency values. And simply limiting port speed for Zaram to value not higher than 5.2G results in nearly perfect upstream. And SCM can be used on router itself in case of Zaram (I tried with OpenWRT and VyOS), and you can achieve similar similar results (depending on choice of SMT).
But downstream is bad for both of them. In fact, on Huawei even worse :)
So, it is general problem, not problem between Zaram and Odido. It a wider issue. It can be masked to certain extent and with different results, and Huawei is masking it so much as it can. But, to be honest, I prefer what I get with hard limit of 5.2G and Zaram. Upload become perfect, latency variation is minimal, no crazy spiking etc.
And Download is tad better than with Huawei in terms of latency variation.
But it looks to me, that on Zaram I never had such crazy values for latency as I see sometimes on Huawei.
So, I think I’ll stick to Zaram, limit TX to 5.2G and see what can I do about RX. (it sucks that on *BSD routers standard mechanisms can’t be used to address this issue at such speeds, and I am not ready to pull trigger on switching from pfSense).
By the way, back to ONTs registration.
Today I also made experiment with cloning REG ID of my Huawei ONT. I cloned it on Zaram, Zaram was able in fact, to pass through authentication (or at least reported that), but it still didn’t work, as then it would stuck in barely responsive state. And, according to the logs, it would keep doing reconnects, as almost immediately after auth it’ll get disconnected. Tried to debug, tried to replace also sn_vsid and sn_vssn (effectively serial number) - same effect.
I’d probably debug more, but debugging something, which responds to commands once every few minutes is a pain :)
And one more update. I recheck. Limiting download bandwidth with fc-codel puts download latency also under control. Also around 5Gbps.
Thus, outcome is simple: stable service is up to 5Gbps. Above that you exchange quality/stability for bandwidth.
And one more update. I recheck. Limiting download bandwidth with fc-codel puts download latency also under control. Also around 5Gbps.
Thus, outcome is simple: stable service is up to 5Gbps. Above that you exchange quality/stability for bandwidth.
Are you on the 8Gbps package? I am having a similar issue as you with the auth, where I get to O5 and then I get reset. If you can access the omci commands, you might be getting the same issue I am. It seems the Huawei OLT’s send some vendor specific command and then alarms due to it.
Yes, exactly same story, reset after O5, when I am trying to mimic Huawei, and proper work when I use Zaram’s ID.
Do you also have telnet becoming PAINFULLY slow at this moment, or it is something local for me?
Yes, exactly same story, reset after O5, when I am trying to mimic Huawei, and proper work when I use Zaram’s ID.
Do you also have telnet becoming PAINFULLY slow at this moment, or it is something local for me?
Oh no mine is fine, but if I try to up the debug level or dumb the omci output to console I loose my telnet and the admin ip becomes unavailable. But I have a MaxLinear chip, from E.C.I. Networks. I am getting 3540 and 373 managed entities commands, from the Huawei OLT. Which my device does not support, then my device gets kicked.
It seems the Huawei OLT’s send some vendor specific command and then alarms due to it.
OLT, depending on detection of ONT, might use different vendor profiles. So, no surprise here. It detects us as huawei
Yes, exactly same story, reset after O5, when I am trying to mimic Huawei, and proper work when I use Zaram’s ID.
Do you also have telnet becoming PAINFULLY slow at this moment, or it is something local for me?
Oh no mine is fine, but if I try to up the debug level or dumb the omci output to console I loose my telnet and the admin ip becomes unavailable. But I have a MaxLinear chip, from E.C.I. Networks. I am getting 3540 and 373 managed entities commands, from the Huawei OLT. Which my device does not support, then my device gets kicked.
I suspect something similar in my case. I just didn’t have enough of patience to debug around non-responsive and regularly lost telnet.. Although it is confusing to me. According to stats, microcontroller pin ONT is used only for 6%. Then I am lost why it becomes sooo slow and, eventually, freeze.
Yes, exactly same story, reset after O5, when I am trying to mimic Huawei, and proper work when I use Zaram’s ID.
Do you also have telnet becoming PAINFULLY slow at this moment, or it is something local for me?
Oh no mine is fine, but if I try to up the debug level or dumb the omci output to console I loose my telnet and the admin ip becomes unavailable. But I have a MaxLinear chip, from E.C.I. Networks. I am getting 3540 and 373 managed entities commands, from the Huawei OLT. Which my device does not support, then my device gets kicked.
I suspect something similar in my case. I just didn’t have enough of patience to debug around non-responsive and regularly lost telnet.. Although it is confusing to me. According to stats, microcontroller pin ONT is used only for 6%. Then I am lost why it becomes sooo slow and, eventually, freeze.
In addition to your bufferbloat, do you see route flipflopping?
Yes, exactly same story, reset after O5, when I am trying to mimic Huawei, and proper work when I use Zaram’s ID.
Do you also have telnet becoming PAINFULLY slow at this moment, or it is something local for me?
Oh no mine is fine, but if I try to up the debug level or dumb the omci output to console I loose my telnet and the admin ip becomes unavailable. But I have a MaxLinear chip, from E.C.I. Networks. I am getting 3540 and 373 managed entities commands, from the Huawei OLT. Which my device does not support, then my device gets kicked.
I suspect something similar in my case. I just didn’t have enough of patience to debug around non-responsive and regularly lost telnet.. Although it is confusing to me. According to stats, microcontroller pin ONT is used only for 6%. Then I am lost why it becomes sooo slow and, eventually, freeze.
In addition to your bufferbloat, do you see route flipflopping?
Ive seen your another thread, yes :) and I think bad routing might be behind bufferbloat also.
traffic looping might heavily increase chance of it
I will check
Is het ook al mogelijk om op het Delta netwerk een eigen ONT te laten registeren? Heb ondertussen een XGS-PON module binnen voor mijn Fritzbox 5590 en zou deze graag willen laten registeren. 😄
+5
Hi
Hi
Inmiddels online met de standaard ONT. De Gideon monteur heeft geprobeerd (samen met back-office) om mijn Zaram online te krijgen, maar dat is helaas niet gelukt. Ik stuur je dus even een PB.
+9
Helaas komt de Zaram niet online. Ik krijg geen DHCP adres. Ik heb je een PB gestuurd.
Ik ben nu online met de Zaram.
FYI voor nieuwe klanten die van KPN/Delta/xProviderx overstappen naar Odido en al eerder XGSPON met eigen ONT gebruikte dient er een “clear” uitgevoerd te worden. Anders blijft de ONT op de oude provider staan in de OLT.
Is er iemand beschikbaar om mijn XGS-PON SN te testen? Ik zou moeten testen met een nieuw serienummer en, als het niet werkt, terugvallen op het originele.
+9
Hey
Enter your E-mail address. We'll send you an e-mail with instructions to reset your password.
