During the past months, the primary focus in the project was on setting up reliable ISDN PRI services via the OCTOI protocol and the related hub ("central office").
Today, we've tested yet another service over OCTOI: IP in HDLC. This broadband technology doesn't use individual 64kbps E1 timeslots like in ISDN, bug it aggregates the entire 31 timeslots of the E1 link (exept for TS0) into one fat HDLC pipe. It is customary to either put raw IP packets, or Ethernet packets, or PPP frames or even Frame_Relay into that HDLC. These method of operation is supported by a variety of routers from the 1990ies.
We've tested two configurations:
Pysical E1 on client side, virtual E1 (dahdi-trunkdev) on hub side¶
In this configuration, an overall inner RTT of 150ms was achieved, a 115ms increase compared to the native underlying IP link.
virtual E1 (dahdi-trunkdev) on both client and hub side¶
In this configuration, an overall inner RTT of 108ms was achieved, a 73ms increase compared to the native underlying IP link.
Yesterday, laforge and manawyrm have managed to establish the first successful calls between two ISDN PBX interconnected via the Proposed_efficient_TDMoIP protocol implemented in osmo-e1d using the icE1usb hardware.
The setup looks as follows:
Using this setup we could validate so far:
- the new, efficient and transparent OCTOI TDMoIP protocol works over public consumer internet access (VDSL on one end / DOCSIS on another end)
- GPS-locked oscillators on both sides provide stable shared clock for cycle-slip free operation
Contrary to other established TDMoIP protocols, it is both transparent / protocol-agnostic and only transmits those timeslots of a E1 PRI circuit that are in use.
We are therefore fairly certain that this setup can be the basis of the larger intended Community_TDMSS7_Network we have in mind. There is lots of work to do, and we appreciate any help (see project issue tracker).
Yesterday, the concepts of the Proposed_efficient_TDMoIP could be validated in practice for the first time using the work-in-progress implementation in the laforge/e1oip branch of osmo-e1d
The setup consisted of two icE1usb USB-E1 interfaces with built-in GPS-DO transporting an E1 line over an intermediate IP network. The clock synchronization has been monitored and sufficient stability over a period of several hours was confirmed, with no underruns/overruns or cycle slips.
The compression feature was working as expected: E1 timeslots with no change compared to the previous frame were supressed. The batching feature (32 E1 frames per UDP packet) was equally working as expected.
There's still a lot of work to do to make this more usable via consume internet connections with NAT and dynamic IP addresses, but the concepts could be shown to work in practice:
- using GPS-DO clocked icE1usb to avoid clock drift
- using batching of E1 frames
- suppressing transmission of timeslots with no data
