OpenBTS Execution Loads and Profile

Running r10019:
  • with partial optimization in L1 and the transceiver
  • with debug messages turned on in L2 and L3

Loads on a 2.33 GHz core 2 duo Apple MacBook Pro are:


num active lines
OpenBTS
transceiver
Asterisk
total

0
0.5%
33%
0%
34%

1
3.3%
37%
1%
41%

2
6.0%
40%
1%
47%

3
7.8%
44%
1.1%
52%

4
11%
46%
1.4%
58%

These loads are percentages of a single core, so a fully loaded system is 200% on this scale.

  • The transceiver is the GMSK radiomodem and the interface to the radio hardware.
  • The OpenBTS application is everything from the L1 TMDA up to the SIP interface.
  • Asterisk runs most of the L3 functions and the transcoders.

In the transceiver, most of this load is in the demodulator. In the OpenBTS application, roughly half of this load is in L1 FEC decoding, with the rest spread around in other components.

These load figures do not include the vocoders, since these are GSM-to-GSM calls and no transcoding is required in the BTS. (The BTS is just forming a conduit for GSM vocoder frames.)

I don't have 7 test phones handy at the moment, but using a simple linear projection, the full-load computational requirement is:


num active lines
OpenBTS
transceiver
Asterisk
total

7
19%
56%
2.5%
76%

... or the equivalent of a single 1.8 GHz core.



 

注:OpenBTS Execution Loads and Profile(原文出处,翻译整理仅供参考!