|
|
||
Vivek Pandey's BlogFebruary 2007 ArchivesJAXWS 2.1 FCS - Fast and FuriousPosted by vivekp on February 02, 2007 at 10:42 AM | Permalink | Comments (9)Phew, what a ride! As usual there has been a few last minutes test failures and re-spinning the build, making sure the Orb is blue in the break area, but finally we are here... JAX-WS 2.1 RI is Final! It was non-stop months of effort by everyone in the JAX-WS team and of course hudson which kept reminding us of any regression we caused before anyone else noticed! JAX-WS 2.1 RI provides convenient programming model built on JAX-WS 2.1 standard APIs and also provides extensions on which middleware developers can build the solutions to enhance it, such as WSIT, Open JBI component, and Glassfish v2. See Kohsuke's blog on extending JAX-WS 2.1 and Jitu's blog on plugging in the Tube and also many developer productivity features. It has been well tested with unit tests, SQE tests and all of the Java compatibility tests(TCK) on top of Glassfish v2 M4 and Tomcat 5.5.x using JDK 1.5_06 and JDK 6. Get the JAX-WS 2.1 RI from here or from maven repository and send your feedback to users@jax-ws.dev.java.net or to the JAX-WS Forum and report your issues and feature requests at IssueTracker. Here is JAX-WS 2.1 RI release announcement:
Performance comparison with Axis2 1.1.1We ran our benchmark tests against Axis2 1.1.1. Details on this benchmark can be found here. People usually care about the server-side performance so the benchmark tests measures the performance on the server side. Large numbers of clients were used to invoke the endpoints simultaneously so we measured number of invocations per second. Since invocations per second varies greatly from tests to tests, we normalized the number so that you'll always see Axis2 as '100%'. On tests like echoVoid where the payload is small, the benchmark tends to reflect the performance of the web service toolkit itself, and there you can see JAX-WS RI 2.1 to be about twice as fast than Axis2. On tests like echoArray where the payload is bigger, the benchmark reflects the data binding performance as well, and here you still see us at least 25% faster. The types of data included empty message (echoVoid), int, float, String, Date, structures of varying sizes(echoStruct) , arrays of varying sizes (echo Array) and binary data of varying sizes (echoSynthetic). You can see in the graph below for simple types you are looking at the performance comparison with the least amount of data binding and for larger messages the comparisons show performance with respect to data binding.
As it looks above in the graph JAX-WS 2.1 RI is performing lot better than Axis2 for all different message sizes. In each of the tests invocations per second includes all the server-side network processing/container etc., so achieving overall 2x performance means that the web services stack has to perform more than double. So get the JAX-WS 2.1 from here or from maven repository and continue providing feedback to users@jax-ws.dev.java.net or to the JAX-WS Forum and report your issues and feature requests at IssueTracker.
| ||
|
|