erstmal ein Hallo in die Runde, ich bin neu hier und verfolge das osmocom-Projekt schon länger. Aktuell bastel ich ein wenig mit Cisco-Equipment und PRI/ATM-Zeugs herum.
Ich habe mir bereits einen funktionierenden Einwahlserver für ISDN (clearmode sip) und Analog gebaut. Allerdings gefallen mir ein paar Dinge nicht.
Zum Einen nutze ich für den Clearmode-Kram auf der Gegenseite einen OneAccess425. Trotz niedrigster Jitter-Buffer komme ich immernoch auf RTTs von >160ms. Niedriger scheint unmöglich zu sein. Einen LANCOM habe ich garnicht erst zum Laufen gebracht. Verbindungen versucht er erst aufzubauen, wenn beide Kanäle gewählt sind, ansonsten kommt direkt ein Fehler.
Gibt es eine Kombination von Geräten, die über SIP Clearmode vernünftige Pingzeiten bei ISDN-PPP-Verbindungen hinbekommen? Würde im zweiten Schritt gerne mit Anderen eine günstige SIP-Clearmode-Geschichte via VPN aufbauen.
Nummer 2: Aktuell gelingt mir eine Einwahl auf der Gegenseite per Datacall nur, wenn ich den SIP-Anruf über eine E1 raus und an einem Cisco 7206 weiterreiche. Der kann Datenanrufe an seinen Ports terminieren, beim 3845 scheint das nicht der Fall zu sein. Oder benötige ich andere Module? (Aktuell die VWIC-2MFT-E1).
Nummer 2b: Gibt es WICs/NMs, welche ISDN (BRI! nicht PRI)-Anrufe annehmen bzw. ein Amt simulieren können, dass dann intern wiederum Daten-PPP-Calls terminiert?
Nummer 3:
Ich habe 2 Digium E1-Karten.
Ist eine reguläre, von DAHDI unterstützte E1-Karte gleichwertig nutzbar wie das osmocom-E1-Interface oder gibt es da fundamentale Unterschiede?
Könnte ich mit den Karten ebenfalls teilnehmen oder benötigt es zwingend das Osmocom E1 Interface?
Und schließlich zum letzten Punkt:
Ich habe mir günstig ein paar MICA-6MOD-Module bestellt. Allerdings sind die zugehörigen NM-xxDM-Module unbezahlbar oder schlicht nicht erhältlich. Hat jemand soetwas zufällig rumfliegen und würde sich davon trennen?
So, erstmal sehr viel Text.
Ich würde mich über ein paar informative Antworten sehr freuen und bedanke mich schoneinmal im Voraus!
zum clearmode kann ich nichts sagen, ich versuche SIP zu vermeiden, wo es irgendwie geht, und verwende lieber TDM-Technik (ggfs. TDMoIP).
Gleichwertig nutzbar wofür? Für ein NT oder TE seitiges PRI interface an einem Asterisk, FreeSwitch, yate o.a.: Ja.
Für die Teilnahme am OCTOI nicht. Die Karte hat keinen GPS-DO und auch sonst keinen nachstellbaren Oszillator. Demnach kann sie nicht snychron zum rest des netzes betrieben werden, und es wird laufend cycle slips, buffer under/overruns und damit verbindungsabbrüche geben.
sorry for the English, doing TDM with a Cisco device can be tricky especially if you don’t have a stable clock source, my wild guess it’s it could be one of the reasons why you get than kind of high latency, I tried doing clearchannel but I wasn’t happy and apart from that I like to keep as much TDM as possible in my setup, but anyway it’s good to check some of this things.
Some indication of timing problems between TDM ports are slip on the controller, you can see that running:
show controllers e1
You can check the clock source running:
show network-clocks
If your clock source it’s “Backplane” it’s basically a free running clock, so you have to figured out how to distribute it, you can recover even recover and distribute the clock from a PBX, like and experiment I did this:
Find your best possible clock source and push it on all your devices!
Question number 2:
It’s not really clear your setup to me, but if you have an E1 Multiflex card configured as “pri-group timeslots” you can’t loopback inside a data call to do dial-in or at least I didn’t find the proper hack yet, officially there is a way for faxes to do so it’s called OnRamp Faxing you can redirect a call into a lua application.
In my setup I’ve got a Cisco 2921 with 2 x VWIC3-4MFT-T1/E1 and 1 x VIC2-2BRI-NT/TE that make me 8 x E1 and 2 x BRI ports in total, this device act as my “main switch” with no SIP or other features. On this “switch” I’ve connected PBXes, RAS, Mediagateway and TDMoIP devices to do the other functions.
Question number 2b:
Yes! VIC2-2BRI-NT/TE you can emulate the network or a terminal and use it for doing call switching between BRI and PRI ports (like the MFT WIC).
If you are interested I can share all the configurations.
danke für den Tipp, das Projekte kannte ich noch garnicht, werde ich auf jedenfall mal testen. Ich vermute aber, die Latenz kommt hauptsächlich vom OneAccess, da der minimum 20ms Jitterbuffer hat, was ja schonmal 40ms gesamt sind. Beim Cisco kannst du (soweit ich weiß) nichts einstellen.
I’m currently just playing around and haven’t 100% understand the differences / config variants, so i settled to pri-group as i can use it for voice/modem and somehow to, at least, passtrough datacalls.
Your configs would be very appreciated, maybe we should connect with calebretro (youtuber) who has a public git as he’s doing similar crazy stuff.
I’ll include SIP for the above mentioned reason as to have an old oneaccess configured for vpn + sip and hand it out to friends to have a virtual isdn-line connected over the internet to my equipment. These boxes are cheap and, despite the ping, work great.
2b:
Can this be done with pure data-calls? (x.75 / ppp) too?
I was only able to terminate voicecalls and route them trough my Auerswald Commander Basic Business-PBX over E1. I’m trying to do this for datacalls also. Would be nice to have somewhat “real” ISDN-ports in my homelab and emulate everything. The PBX is nice, but lacks certain features as it is tailored to some specific usecased and not very flexible.
The definition of “crazy stuff” here is on a different level
Briefly:
pri-group use ISDN signalling you need to configure the voice feature and have the dsp
ds0-group use CAS signalling in different flavors depending of the configuration the dsp are required (for example if you use a timeslot into a dialer)
channel-group use CAS signaling but you can configure specific bandwidth for a timeslot
tdm-group is to build an add-drop multiplexer between channels and ports
You can mix ds0-group and channel-group on the same port, but are not compatible with pri-group or tdm-group; there are a lot of obscure combinations of config for make works others signalling like CoreNet / CoreNet-NQ.
Using pri-group with protocol-emulate network and isdn switch-type basic-net3/primary-net5 is the closest thing to a basic isdn switch you can do, there is a really annoying thing and is you have to provision the dsp for every single timeslot you configure.
There are some things I can’t figured out, for example call progress over BRI interfaces on Cisco are wrong I can hear the wrong call progress tone (I get the dial tone instead the ringing tone), same configuration over the E1 ports just works out of the box. We’ve some discussion about it over IRC and one hypothesis is the firmware/hw revision of the VWIC.
The dialplan is really convoluted because I tried to emulate a proper network so are some translation the prefix, type and plan, furthermore I used a bunch of “destination-pattern” dial-peer instead of working properly with the “incoming called-number” this is wrong, and I planned to fix it in version 3 when I have some time to spare.
2b:
Yes you can process data call passing trough, for some weird reason, during the call progress, one dsp slot is used until the call is established, but it is what it is.
Most of us use Auerswald PBXes I guess for the same reason: to get a bunch of S0 ports for cheap. Yes, we agree those are really basic.
Make sure the Subscriber → Telephone numbers → Slot X XS0 the Device Type is “ISDN-PC-contr.” if you look at the wiki there are some notes regarding this issue, and as I mentioned before check for the clock configuration, you can either use the PBX (there is a jumper on the S2M linecard) or the Cisco (under each E1 controller “clock source”) as a master clock. You have to try what give you the less time slips.
I’ve tested this specific call flow with success, it’s extremely stable I had data call up for weeks at the time: Modem --S0--> PBX --E1--> Cisco --E1--> PBX --S0--> RAS
Yesterday i finally had time to play a bit more. Installed a NM-1CE1T1-PRI, which is back-to-back with a VWIC-2MFT-E1.
Now i am finally able to make X75-calls either from SIP (slow ping), E1 or directly from a VIC2-BRI-NT/TE! (which is blazingly fast regards to the connection sequence.)
Takes less than a second for my XP-machine to dial and get a MPPP bundle of 2 channels! Ping is around 23ms. Great.
I’m currently trying to get a NM-xxDM-carrier from some seller but no mather what i try, either they just cancel the order and raise the price afterwards or reject my offer, they all think that the obsolete modems are worth gold.
Not talking about the PVDM2-DM-cards … who is willing to pay those huge amounts?!
It’s with an external fritz!card usb in a XP-VM running on proxmox.
I bet with internal card (isa or pci) the ping would be even lower.
Another test i’m doing right not is using a Zyxel Prestige 400.
With 4 channels (2 isdn bri lines) bundled i get around 29kb/s and a ping of 38-40ms. The Zyxel-CPU is adding some overhead.
Will test with internal cards the next days when i got some time …
EDIT: without NAT and real routing, the zyxel-quad-bond gets around 26ms RTT. Nice!
thinking about my discoveries makes me wonder what latency i could expect from cisco-2-cisco isdn over sip over ipsec… and i also have an urge to have a dedicated E1 line for my retro workstation next to my regular business workstation. cisco 1921 order placed! will report back!
Clearmode / clear-channel stinks.
Zyxel-Router → ISDN NT/TE [Cisco 2921] SIP over T3 → T3 [Cisco 3845] → E1
around 200mS RTT.
Will test with a 1921 and a OneAccess via S.GHDSL too, but i bet the performance will be similar.
Problem with clearchannel seems to be the jitterbuffer which anoyingly cannot be changed on that type of gear.
