[omniORB] Optimizing the roundtrip-time of function calls?

[Duncan Grisby]

On Saturday 25 April, Oliver Frick wrote:

> How can I optimize a function call's roundtrip time?
> I have measured the up-call and return time of a corba
> call between two processes on the same host.
> The result is always that the up-call time is much smaller 
> than the return time. The ratio is typically 1/5, 
> sometimes 1/100 or more.
> The function is rather simple, it has no arguments
> (or two longs; I added them for testing) and always two
> longs in a returned struct.
> 
> If the ORB is able to up-call the other object in, say 
> 150 usec, why does it need 1.5 msec to return? Can I 
> reduce the return time by configuring omniorb?

How are you measuring it?  I would not expect that, and have never seen
anything to suggest that return handling was slower than incoming call
handling.

How are you configuring omniORB?  Are you using bidirectional GIOP?

Cheers,

Duncan.

-- 
 -- Duncan Grisby         --
  -- duncan at grisby.org     --
   -- http://www.grisby.org --