Firstly remember that you're probably not comparing apples to oranges.
The default in JMS is to use persistence; that means the messages are
all written to disk before they are sent (on both the request and
probably response) which is gonna add hugely to the latency as each
thread is gonna be spending most of its time waiting for stuff to be
stored on disk. So if you really wanna compare like for like, turn off
persistence in JMS - or use HTTP with WS-RM with a *persistent*
implementation (not a RAM version).
Also make absolutely sure you are pooling all the JMS resources;
creating connections, sessions, producers, consumers on a per message
basis is insanely slow - they should all be created up front.
For more help see...
http://activemq.apache.org/how-do-i-use-jms-efficiently.html
http://open.iona.com/wiki/display/ProdInfo/FUSE+Message+Broker+Performance+Tuning+Guide
But hey if you're happy with HTTP and don't have any reliable
messaging requirements there's no real need to change.
On 26/11/2007, Mayank Thakore [EMAIL PROTECTED] wrote:
Hi Ulhas,
I am using default temp. queue. Earlier when working with XFire we had
tried static reply queue. But are just gearing up with CXF.
I was interested in relative performance of http / jms. We have been
given performance requirements for our project. While http meets just
as it is, jms is a far cry from what we must achieve.
I am currently looking at three things to ascertain where jms could be
going slow:
1. OpenJMS server (persistence causing delay?)
2. Our WSDL (sync/async differs?)
3. CXF jms transport
I am still reading what a web service is after finishing my first
project where we had used XFire.
Thanks for any and all help
Regards
Mayank
On Nov 26, 2007 9:13 PM, Ulhas Bhole [EMAIL PROTECTED] wrote:
Hi Mayank,
I don't think anyone has done the benchmark on the cost of
request/response on JMS transport.
Are you using static reply queue or using default temp. queue for response?
Regards,
Ulhas Bhole
Mayank Thakore wrote:
Hi,
I wrote a simple WSDL based web service.
In HTTP I am getting 1000 request/response in 19 secs.
But, in JMS 1000 request/response cost 110+ secs.
Is this OK? Are there any benchmarks?
Any pointers on what can be done for JMS?
I am using CXF 2.0.2 and OpenJMS.
Thanks!
Mayank
IONA Technologies PLC (registered in Ireland)
Registered Number: 171387
Registered Address: The IONA Building, Shelbourne Road, Dublin 4, Ireland
--
James
---
http://macstrac.blogspot.com/
Open Source Integration
http://open.iona.com