RH> Option 1:  

 RH>   process A                                   process B   
 RH>   =========                                   =========  
 RH>   thread A1  --->  named pipe instance 1 ---> thread B1  -\
 RH>   thread A2  --->  named pipe instance 2 ---> thread B2  ---\ 
 RH>     ...                                                  -----> queue
 RH>     ...                                                  ---/ 
 RH>   thread An  --->  named pipe instance n ---> thread Bn  -/          


 RH> Option 2:

 RH>   thread A1  -\            
 RH>   thread A2  ---\   
 RH>     ...      -----> queue ---> single instance named pipe
---> process B
 RH>     ...      ---/         
 RH>   thread An  -/            

 RH> I am interested in which case I obtain best 
 RH> performance OF PROCESS A (I am less intersted in the 
 RH> perfomance of process B).

Then I would choose option 2,  since queues are a lot faster (and use less
[lan] resources) than named pipes.

 RH> With option 2 process A has both queue and pipe 
 RH> processing. So I expect the process more resource 
 RH> consuming. But queue processing might be lighter than

But the overall number of pipes is less (one vs many).

 RH> Please consider in your answer that the processes may run on the same or 
 RH> different systems. 

I figured as much -- if it was always to be on the same computer, you
wouldn't want to use a named pipe at all.


--- Maximus/2 2.01