[rabbitmq-discuss] Steady performance degradation

Matthias Radestock matthias at lshift.net
Sat Sep 26 05:26:29 BST 2009


Matthias Radestock wrote:
> I can't be sure of anything, given that I haven't seen the complete 
> code. But the order of suspects is: the application code, the 
> system/network setup, the Java client lib, the server.

Also, the following would be on my list of things to check:

1) Is the test exhausting some server resources, in particular CPU or 
network bandwidth?

2) Is the measurement code correct?

3) You mentioned the use of a basic.qos prefetch limit - do you get the 
same results w/o that?

4) Are messages marked as persistent? If so, do you get the same results 
w/o that?

5) Are you using tx at all? If so, do you get the same results w/o that?

I would also recommend that you simplify your test code as much as 
possible. If you can condense it to something along the lines of the 
shipped examples then a) you can send it to us, and b) it will be much 
easier to track down where the problem may be.



More information about the rabbitmq-discuss mailing list