Vibhor
Generally speaking we test web services the same way we test HTTP
systems. We use a client load driver (such as apachebench, faban,
JMeter, Mercury LoadRunner etc) on a very fast local network.
But - you will say - doesn't this completely ignore network delays and
effects? Yes it does. The first aim is to ensure that the service can
handle the load independent of network latency and delays.
There are ways you can insert latency - for example the tcpmon tool
from Apache WS can add random delays. But in general its more useful
to test the actual scenario - i.e. how long does it take user X from
their own network. And of course you can make approximations based on
knowledge of timings from ping.
Paul
PS did you ever find out what was happening in your previous post?
On 6/11/07, Vibhor_Sharma <[EMAIL PROTECTED]> wrote:
Hi Experts
I wanted to know what are the recommended ways of having or doing
performance benchmarking for web services. Normally web services can be
tested within your limited testing environments but the moment you expose
them they can be consumed in any manner. You have no control over the
network latency and other parameters that control the internet. So how do we
give recomendations to our customers that this is what they can achieve in a
given environment.
Is there any ways we can introduce network latency and give different
statistics?
We have used SoapUI, Jmeter to perform load testing. We have also tried to
use the standalone Java test clients, C# clients to achieve the results.
We are using Axis2 1.1.1 for the web services engine and they are deployed
on Jboss 4.0.5 GA on Linux (REDHAT) boxes.
Thanks
Vibhor
--
Paul Fremantle
Co-Founder and VP of Technical Sales, WSO2
OASIS WS-RX TC Co-chair
blog: http://pzf.fremantle.org
[EMAIL PROTECTED]
"Oxygenating the Web Service Platform", www.wso2.com
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
