When I run benchmark (locally) on static resources, GF 3.1.1
(out-of-the-box) shows better results than JBoss 7.1.0 (out-of-the-box).
I deployed webapp, which contained 18839 bytes txt file and ran ab
ab -n 100000 -c 20
http://localhost:8080/myapp/test.txt
once to warm up the servers and then 3 more times to get results.
Here they are:
GF:
Transfer rate: 170703.49 [Kbytes/sec] received
Transfer rate: 170194.09 [Kbytes/sec] received
Transfer rate: 171569.14 [Kbytes/sec] received
JBoss:
Transfer rate: 168732.30 [Kbytes/sec] received
Transfer rate: 152857.97 [Kbytes/sec] received
Transfer rate: 166922.91 [Kbytes/sec] received
Don't want to speculate which server is better or worse, I can just say
that results are comparable and we can not talk about *significantly*
better results of one or another.
Regarding jsp - looks like you're right, my tests also show GF performs
worse than JBoss (both out-of-the-box), may be it's just out-of-the-box
configuration difference or some kind of issue... we're looking into this.
On 03/11/2012 04:14 PM, Radim Kolar wrote:
> It looks like Tomcat bug most likely because both jboss and glassfish
> are using Tomcat as servlet container.
Another common pieces are your webapp and environment config ;) I
wouldn't exclude that. I wasn't able to reproduce truncate problem even
w/ webapp you sent me directly (btw thank you for that).
Thanks.
WBR,
Alexey.