Colocated Hub » History » Version 3
laforge, 04/21/2022 06:22 PM
1 | 2 | laforge | {{>toc}} |
---|---|---|---|
2 | |||
3 | 1 | laforge | h1. Colocated Hub |
4 | |||
5 | 2 | laforge | This page exists to collect some planning about the future colocated OCTOI hub, see #5542 |
6 | |||
7 | h2. Physical Setup |
||
8 | |||
9 | h3. Server (specs) |
||
10 | |||
11 | High level: |
||
12 | * Some 2U machine with capacity to host preferably >=2 full-height PCIe cards (TE820). |
||
13 | * SRV-IO to map TE820 into a VM for easier debugging of potential DAHDI kernel crashes |
||
14 | * amount of RAM is not very important, the use case has no high RAM requirements |
||
15 | * CPU is more critical given all of the the byte-shifting and lots of copying / context switches we have to do for each trunk in the current software architecture |
||
16 | * storage is also not very critical, other than for log files / monitoring |
||
17 | |||
18 | Candidates: |
||
19 | * @laforge has experience with Dell PowerEdge R740 (up to 3 full-height slots) |
||
20 | * @gruetzkopf has suggested HP DL830p G8 (up to 6 full-height slots) |
||
21 | * @laforge is checking SuperMicro options with riser cards for PCIe |
||
22 | |||
23 | |||
24 | h3. DAHDI cards |
||
25 | |||
26 | @laforge has at least two TE820 that can be used for this purpose |
||
27 | |||
28 | |||
29 | h3. icE1usb |
||
30 | |||
31 | We need at leaset one icE1usb as a GPS-disciplined clock master. |
||
32 | |||
33 | We can also use that one to compare physical (icE1usb+TE820) setups against DAHDI-trunkdev. |
||
34 | |||
35 | We could deploy 1-2 more, just in case. |
||
36 | |||
37 | |||
38 | h3. RAS / dial-up server |
||
39 | |||
40 | We should install at least one RAS server next to it, for dial-up modem / isdn termination |
||
41 | |||
42 | 3 | laforge | * Definitely the Livingston Pormaster3 currently next to the hub. Needs 1xE1 on the Server |
43 | * Likely also a Cisco ASR5400 @laforge has acquired (much higher capacity, likely more featurs) |
||
44 | 2 | laforge | |
45 | |||
46 | h2. Software |
||
47 | 1 | laforge | |
48 | * Like in the current hub at @laforge's basement, we should run the entire yate / osmo-e1d / dahdi-trunkdev inside a KVM with SRV-IO mapping the PCI devices |
||
49 | 3 | laforge | |
50 | |||
51 | h2. Misc / TODO |
||
52 | |||
53 | h3. local verification with modem/ISDN-TA |
||
54 | |||
55 | It would be great if we could somehow verify local calls via the yate / hub excluding the complexity of the TDMoIP to compare calls without TDMoIP to calls with TDMoIP. |
||
56 | |||
57 | Unfortuantely I'm not aware of any really small PBX that would terminate the E1/S2M and offer S0/POTS ports. Adding antoher 3U for a PBX seems excessive. |
||
58 | |||
59 | Doing calls between PM3 and ASR5400 is of course an option, but ideally we'd have a real physical modem attached via a serial port, and a real HFC-S-USB for test purpose. |