Subject: Re: [linux-audio-dev] Distributed audio processing
From: Benno Senoner (sbenno_AT_gardena.net)
Date: Fri Oct 06 2000 - 11:53:25 EEST
Perhaps I was not clear enough in my previous mail:
I am speaking mainly about streaming low latency audio streams (serveral)
between multiple boxes.
I don't think that MIDI is a big problem because it uses much less bandwidth
than audio (and from a 100Mbit LAN POV, almost zero :-)
= 0.03% of the available bandwidth)
but of course it would be annoying to see several MIDI events delayed by
10-30msecs every now and then.
Anyway what do you mean with "our LAN here":
"our idle LAN" or our "our busy LAN" ?
Plus is the LAN switched or not ?
Think about the fact that in distributed audio processing apps I have in mind,
you assume that you do not have to share the net with others, which can
cause high latency peaks.
Benno.
On Thu, 05 Oct 2000, John Lazzaro wrote:
> > So I was wondering if one could achieve worst case latencies of 5-10msec
> > on a 100Mbit Lan.
>
> I don't know enough to say for sure, honestly ... our own LAN here in CS
> at Berkeley gives you "nominal" sub-1ms round-trip times, with 3-sigma
> bumps into 3-7ms category -- but 3-sigma isn't "worst-case" by far. A
> sensible Forward Error Correction strategy customized to MIDI can
> certainly help, as can other things ... but its probably better to think
> about the problem in terms of "how long between bad situations, on average"
> and not "absolute guarantees" -- after all, the "All Notes Off" command
> in MIDI-on-a-wire shows that even in the hard-cabled serial-link world,
> its good to have fallback for when the guarantee dails :-).
>
> -------------------------------------------------------------------------
> John Lazzaro -- Research Specialist -- CS Division -- EECS -- UC Berkeley
> lazzaro [at] cs [dot] berkeley [dot] edu www.cs.berkeley.edu/~lazzaro
> -------------------------------------------------------------------------
This archive was generated by hypermail 2b28 : Fri Oct 06 2000 - 10:23:34 EEST