Casual Test on Remoting Protocols

 I’ve designed two scenarios to investigate three remoting protocols that

said to have good performance, XML protocols are excluded.

 

XML protocols are born to be slower than the native RMI and some light weight http based protocols like HttpInvoker and Hessian, etc.

 

 It seems HttpInvoker has a bit performance advantage over RMI, the tests are  carried out on two different hosts in a LAN network. The server side is a 1.8GHz CPU, 1G Memory PC, the client side is a 1.8GHz, 1.5G memory notebook.

 

Scenario 1:

  It simulates the publish process of app which transfers lots of Objects.

 

Total Requests processes

1minute/10k Objects	RMI	HttpInvoker	Hessian
10Con-Requests	2039	2111	809
20Con-Requests	2376	2543	864
30Con-Requests	2855	2840	924
50Con-Requests	3207	3079	976
100Con-Requests	3581	3832	1260

 

 

Average Response Time

1minute/10k Objects	RMI	HttpInvoker	Hessian
10Con-Requests	0.288	0.278	0.772
20Con-Requests	0.438	0.473	1.509
30Con-Requests	0.649	0.666	2.188
50Con-Requests	0.968	1.008	3.720
100Con-Requests	1.731	1.602	6.569

 

 

 

Scenario 2:

  It simulates business logic which spends 500 miliseconds.

 

Total Requests processes

1minute/10k Objects	RMI	HttpInvoker	Hessian
10Con-Requests	1000	1141	221
20Con-Requests	1889	2267	258
30Con-Requests	299	3407	289
50Con-Requests	4743	5472	369
100Con-Requests	4743	10389	538

 

 

Average Response Time

1minute/10k Objects	RMI	HttpInvoker	Hessian
10Con-Requests	0.592	0.522	6.012
20Con-Requests	0.625	0.522	6.012
30Con-Requests	0.589	0.521	8.907
50Con-Requests	0.62	0.542	14.086
100Con-Requests	0.620	0.558	28.275

 

 

 

 

Resources:

1) Middleware remoting protocol migration              -------http://blog.nominet.org.uk/tech/2007/03/13/middleware-remoting-protocol-migration/

2) http://www.devx.com/Java/Article/27300/1954  -------New Protocol Offers Simple, Efficient Java RMI

3) Java Remoting: Protocol Benchmarks           -      ------http://q.sohu.com/forum/5/topic/1148909