Socket Performanve vs Direct communication
- From: "Peggy" <pawanarya@xxxxxxx>
- Date: 23 Jul 2006 20:01:48 -0700
Hi
While performance/stress testing my application (a COM+ Server
component) we drive the transaction to the component via a Socket
Server app, listening to a port and passing the incomming transactions
to the COM+ application. The Socket App then returns the response back
to the opened socket to the client. For a 100 users scenario, 8 CPU box
I am able to get 130 Transactions per seconds.
How ever if I try to pass these transactions directly to an instance of
the COM+ component without the Socket Server, I get performance f 190
Transactions per seconds.
Does spawning a Socket Server really incurr that kind of an overhead,
in this case it is 30%, or I am doing smething wrong with the Socket
Server app. It runs Async and spawns a new thread for each new
connection.
Thanks in Advance
Pawan
.
- Follow-Ups:
- Re: Socket Performanve vs Direct communication
- From: Barry Kelly
- Re: Socket Performanve vs Direct communication
- Prev by Date: Re: Profiling and inlining
- Next by Date: Re: Memory Cleanup for Hashtables
- Previous by thread: Re: good tool for dev stress testing
- Next by thread: Re: Socket Performanve vs Direct communication
- Index(es):
Relevant Pages
|