HTTP Tunneling performance
0 posts in topic
Flat View  Flat View
TOPIC ACTIONS:
 

Posted By:   Paul_Atkins
Posted On:   Friday, January 4, 2002 09:38 AM

I have a (private) test network setup which has dedicated server and client firewall machines. The client firewall machine is also running a proxy. The RMI server exports the objects on a known port which has been opened in the server firewall. The client firewall does not allow direct connection, and so the RMI client is forced to tunnel. It succeeds in this using the HttpToPort factory and the remote calls execute properly. However, it appears that when tunneling there is about a 2.4 second overhead per remote call compared to the direct connection. Strangely though an occasional call (1 in 30?) goes through much quicker, about 10 times faster. Switching on rmi transport debug (-D   More>>

I have a (private) test network setup which has dedicated server and client firewall machines. The client firewall machine is also running a proxy.

The RMI server exports the objects on a known port which has been opened in the server firewall.

The client firewall does not allow direct connection, and so the RMI client is forced to tunnel. It succeeds in this using the HttpToPort factory and the remote calls execute properly.

However, it appears that when tunneling there is about a 2.4 second overhead per remote call compared to the direct connection. Strangely though an occasional call (1 in 30?) goes through much quicker, about 10 times faster.

Switching on rmi transport debug (-Drmi.transport.proxy.logLevel=1000) I see that when tunneling it creates a new socket for every call (this is not the case on the direct connection). Surely this can not account for such a long delay.


Has anyone else seen this behaviour - is this performance typical or am I missing some obvious configuration tweak?

   <<Less
About | Sitemap | Contact