[omniORB] ORB performance benchmark

R.vd.Leek@fokkerspace.nl R.vd.Leek@fokkerspace.nl
Tue, 5 Sep 2000 13:13:53 +0200


	Hello list,

	I'm doing some basic research on the performance of
	freeware (GPL-like licensed) ORBs. Therefore I need some 
	simple ORB performance benchmarks.
	The performance tests and their results are viewed from 
	an application designer point of view, so I'm looking at 
	practical things here.

	My first test would be to measure the object method 
	invocation time (round trip time) were the server and 
	client run in separate memory address space on the same 
	machine. 
	IMHO this would give a reasonable estimation on the 
	performance of the ORB's marshalling and dispatching 
	procedures.

	In an example application the client would instantiate X
	MiTest objects (see interface displayed below) and call 
	the ping method from each object in the order they were 
	created.
	This test (method invocation) could be repeated Y times.

	module Benchmark {
		interface MiTest {
			void ping ();
		};
	};

	- Does proxy caching mechanisms (all objects implement the 
	same interface) influence the tests results?
	- Can you recommend improvements on this benchmark which 
   make the test more representative for the comparison of 
   ORB performance?
   
   TIA, rob.
--
Rob van der Leek
Fokker Space b.v.
Remote Sensing Group
e-mail: r.vd.leek@fokkerspace.nl
tel.: +31-(0)71-5245531