[sdiy] MIDI over Ethernet

[Martin Naef]

Magnus,

Magnus Danielson wrote:
> Yes and no. Depends on your timing requirements. If you run two Ethernet ports
> back-to-back you have fairly good timing, until you add additional traffic as
> it hits the play-out buffers. If you run through a switch, the other traffic
> from other ports join in and helps to unstabilize the timing. Multiple hops of
> Ethernet switches... it gets worse again. Even if there is no other traffic,
> timing gets screwed up by a number of anisochronous and asynchronous mechanism
> that contribute here and there on timing. May not be a sindfull load, but it
> adds up.

All true - but has anybody recently *measured* how much these things add
up in a local network?

9 years ago, I've done some extensive benchmarking of network latencies
for supercomputing clusters, both using specialised networking hardware
as well as standard switched ethernet. Point to point latency using MPI
over Ethernet (built on top of TCP - it *had* to be reliable) sure
showed some significant variability, but with an average packet latency
of around 120us, it was still heads and shoulders above MIDI, which
takes about a ms to transmit a *single* message, whereas with UDP you
can easily pack a hundred of them into a single packet... Unfortunately,
I can't remember the worst case figures, but I think it was still way
below what you might get with an 8x8 midi interface driven through an
RS232 serial port, which people found perfectly acceptable for a long time.

I'd be interested in hearing about more up-to-date measurements.

Bye
Martin

-- 
http://www.navisto.ch
http://www.myspace.com/navisto