Kannaiyan Natesan wrote: > I installed debian etch with asterisk in net 5501. The installation is > really awesome. Went without even a single problem. > > But now the codec translation time is high. Looks like it is the > software performance issue. > Astlinux is nicely optimized from this view. > > Here is the translation table. > > kanspbx*CLI> show translation > Translation times between formats (in milliseconds) > Source Format (Rows) Destination Format(Columns) > > g723 gsm ulaw alaw g726 adpcm slin lpc10 g729 speex ilbc > g723 - - - - - - - - - - - > gsm - - 11 11 27 11 10 44 - 294 - > ulaw - 42 - 1 18 2 1 35 - 285 - > alaw - 42 1 - 18 2 1 35 - 285 - > g726 - 57 17 17 - 17 16 50 - 300 - > adpcm - 42 2 2 18 - 1 35 - 285 - > slin - 41 1 1 17 1 - 34 - 284 - > lpc10 - 61 21 21 37 21 20 - - 304 - > g729 - - - - - - - - - - - > speex - 73 33 33 49 33 32 66 - - - > ilbc - - - - - - - - - - - > > Anyone have any tips to optimize it, please do let me know.
Kannaiyan, That's interesting. Did you install binary executables, or did you build from source code, with the processor type set up for the 5501? Transcoding between compressed codecs involves heavy DSP processing, so a low-power CPU needs all the help it can get. The CPU-specific compiler optimizations might not make much difference, but are worth trying (if you haven't done so already). For interest, I just found a Cisco white paper which states: "The generally-accepted limit for good-quality voice connection delay is 200 milliseconds one-way (or 250 ms as a limit). As delays rise over this figure, talkers and listeners become un-synchronized, and often they speak at the same time, or both wait for the other to speak." If you want to transcode to GSM on 4 simultaneous calls, then the processing delay is multiplied by 4, so you use up the whole 200 millisecond end-to-end delay budget just for transcoding at your end. So perhaps you need a faster CPU (or more bandwidth to give more choice of codec). - Martin. _______________________________________________ Soekris-tech mailing list [email protected] http://lists.soekris.com/mailman/listinfo/soekris-tech
