6 Replies Latest reply on Oct 31, 2013 6:53 AM by swd847

    undertow from wildfly-8.0.0.beta1 performance

    ivagulin

      Today I wrote a small web app to test undertow performance improvements against jboss web. Application expose one jpa entity through rest. To my surprise wildfly-8.0.0.beta1 performs 2 times slower than jboss-as-7.1. Is it expected behaviour at this stage or I hit some performance issue?

       

      I used weighttp for proper keep-alive and non-blocking client io. It requires some patching to work with wildfly(attached). As I understand main problem is weighttp make 2 socket read calls on undertow: 1st for http headers, 2nd for chunk length and data. On jboss-7.1 first socket read whole data: both headers and chunk.

       

      jboss web
      [igor@ivagulin-pc ~]$ weighttp -c1 -t1 -n100000 -k http://127.0.0.1:8080/j2ee-domains/rest/domains/2/
      finished in 10 sec, 395 millisec and 680 microsec, 9619 req/s, 2789 kbyte/s
      procs -----------memory---------- ---swap-- -----io---- -system-- ----cpu----
      r  b   swpd   free   buff  cache   si   so    bi    bo   in   cs us sy id wa
      1  0      0 12508356  99340 2494828    0    0     0    88 30480 58392 10  5 86  0

       

      undertow
      [igor@ivagulin-pc ~]$ weighttp -c1 -t1 -n100000 -k http://127.0.0.1:8080/j2ee-domains/rest/domains/2/
      finished in 20 sec, 371 millisec and 759 microsec, 4908 req/s, 1231 kbyte/s
      r  b   swpd   free   buff  cache   si   so    bi    bo   in   cs us sy id wa
      0  0      0 12553192  99492 2495536    0    0     0     0 22264 42081 11  3 86  0