There is a presentation from 2008 by Vincenzo Nicosia describing work on HRT erlang (HARTE), which also describes the problems regarding hard real-time of the current BEAM, etc.
Have a look at http://www.erlang.org/workshop/2007/proceedings/05nicosi.pdf
0
bdonlan
On
Garbage collection can lead to pauses of unpredictable duration; as such, you can't put hard upper bounds on latency. The definition of hard real time is basically that you can set hard bounds on latency, so this is a problem.
0
I GIVE CRAP ANSWERS
On
Oh, nobody has answered why it is good for soft realtime though:
In Erlang, each process has its own garbage collector. So if the processes each are kept with a rather small heap, then GC pauses tend to be very small and this gives a really good realtime behaviour.
There is a presentation from 2008 by Vincenzo Nicosia describing work on HRT erlang (HARTE), which also describes the problems regarding hard real-time of the current BEAM, etc. Have a look at http://www.erlang.org/workshop/2007/proceedings/05nicosi.pdf