General question about concurrent connections and Netty.

javadevmtl java.dev.mtl at gmail.com
Tue Jun 15 20:35:23 EDT 2010


What causes a connection reset by peer on the client side?

So here is some more info from my testing. I just want to clear things up
and make sure it's not Netty or my application because there is finger
pointing lol. One side is blaming the server, one side the client.

there are 2 test applications
- Jmeter (latest Version)
- Custom tool created by my QA team

When using JMeter, the server including full business logic can handle up to
about 280 requests/sec. JMeter starts spitting the errors mentioned in the
previous thread at 280+ requests/sec.

When using the custom QA tool randomly (sometimes it works perfect sometimes
it doesn't) at about 150 requests/sec the client test tool starts to spit
out connection reset by peer.

My stance is that JMeter is programmed better and more efficient and I need
to prove this to the team.

JMeter is configured to use 50 threads "users" and each user is responsible
for creating as many connections as it can when it can as fast as it can.

The custom QA tool creates 200 threads per second 1 connection per thread.
Every 1 second there is 200 new threads firing. The connection reset by peer
only happens when the average request/response is close or above 1 second.
So that said obviously the concurrent connections open within 1 second are
more.

Also i noticed, with Jmeter the average request/response time is always
about 50-200ms. While with the QA test tool it randomly goes from 50-1500ms.
Each subsequent test yields different response time while with JMeter it's
constantly the same average.







-- 
View this message in context: http://netty-forums-and-mailing-lists.685743.n2.nabble.com/General-question-about-concurrent-connections-and-Netty-tp5179056p5184550.html
Sent from the Netty User Group mailing list archive at Nabble.com.


More information about the netty-users mailing list